Oral evidence: Hate Crime and its Violent Consequences, HC 609

Oral evidence, 14 March 2017.

Witnesses

Peter Barron, Simon Milner and Nick Pickles.

Chair

May I welcome you to our Committee session today and our inquiry into hate crime? I put on the record my declared interest, as a member of the all-party group, Reclaim the Internet. I welcome our witnesses from Twitter, Facebook and YouTube. We note, as a Committee, that we are able to bring you to give evidence because you have staff in the UK and community standards, whereas some of your counterpart social media organisations unfortunately have neither community standards nor staff in the UK for us to ask to appear before us. We recognise that you do more than many social media companies to address issues around hate crime. However, we also recognise that you are the biggest companies and have, between you, millions of UK users on your platforms every day. We want to ask you about your response and the way in which you deal with issues around hate crime and abusive content on your platforms. We have sent each of you some individual examples of areas and content that we have concerns about. I want to start by asking you about some of those, before we get on to the general points. Let me start with YouTube, Mr Barron. As part of your community standards, you say, “we don't permit hate speech”, which is “content that promotes violence or hatred against individuals or groups based on certain attributes”, such as race, religion and gender. You also say that it is “not acceptable to post malicious, hateful comments about a group of people solely based on their race.” And yet, until yesterday, when we raised concerns with you, you still had a whole load of neo-Nazi videos on YouTube. Why?

Peter Barron

There are three videos that you sent us. One related to David Duke and two related to organisations in this country, including National Action. Of those three videos, we have removed two. One was on the grounds of our hate speech policy, because it was clearly condoning or promoting violence against a particular group, so that was removed. In another case, the video related to National Action, which is a proscribed organisation; it was removed because it was representing a proscribed organisation and was therefore illegal content. Whether the content of the video itself, which was a rally, would have counted as hate speech is uncertain. The third video related to David Duke. I am not here to defend the content of the video; it was anti-Semitic, deeply offensive and shocking. Our teams looked at all three videos in exactly the same way that they would look at content that is flagged within our systems. I saw the email debate that went on among our legal and policy teams to make a decision about that video. The decision they took was that that video should stay up. David Duke is obviously a well-known figure and has made many videos in this area, and we of course have removed many videos that do veer into hate speech and promoting violence against particular groups. In this particular case, the judgment of our team was that the video did not cross the line into hate speech. It was extremely offensive and shocking material, but it did not veer into the category of hate speech and therefore stayed on the platform.

Chair

Okay. Let’s come back to that one. Let me just deal with the video from National Action. Why did you not take it down before?

Peter Barron

The way that our system works is on notice and takedown. We rely on our community of more than 1 billion people around the world to flag videos to us that they think may break our guidelines. On each video on YouTube, you have a flag, with a little drop-down menu that shows the different elements under which it may break our guidelines.

Chair

But National Action is proscribed by the Government. Not many organisations are proscribed by the Government. You did not have any system in place to ensure that you, as an organisation, were taking off your platform content that had been set up there by an organisation that was proscribed for terrorist links by the Government.

Peter Barron

When drawn to our attention, that material comes down very quickly. We get something like 200,000 flags a day, and we are hitting a target of about 98% reviewed within 24 hours. When drawn to our attention, our teams look at that material. If it breaks our guidelines, it comes down. That is what happened in this case. You drew it to our attention. Things are flagged to us by a variety of means, including through the platform. You flagged it to our attention. We looked at it and within a couple of hours took action on it.

Chair

That suggests we have to do an awful lot of inquiries to get you to just take down some basic stuff from a proscribed organisation.

Peter Barron

No, I am not making that point. We work with the community on YouTube of 1 billion people. We also work with trusted flaggers. We have experts out there in the community. The CTIRU, for example, are trusted flaggers, and they are extremely qualified experts in this area. They are constantly looking for terror-related videos and flag them to us. We are in constant touch with them—maybe every other week we hear from them. They are able to flag content in bulk to us, and their accuracy rate is around 80%, which is a very high degree of accuracy compared with lower rates in the community more broadly.

Chair

But there are not that many proscribed organisations. Do you not even feel any sense of responsibility, as a multibillion pound organisation, to at least check that you are not distributing material from proscribed organisations?

Peter Barron

We have 400 hours of video uploaded on to YouTube every minute, which is an extraordinary amount of content. Clearly we do not want illegal content on our platforms, and when flagged to us we remove that as quickly as we possibly can. We are hitting targets to remove content within 24 hours in 98% of cases.

Chair

I will take that as a no—you don’t feel you have any responsibility to do your own searches, even of proscribed organisations.

Peter Barron

We have a responsibility to act on notice and takedown.

Chair

Let me ask you about the video called “Jews admit organising white genocide”. In what circumstances is “Jews admit organising white genocide” not a statement of hate speech?

Peter Barron

There is no clear definition of hate speech in British law. We have our own guidelines around hate speech. The guideline that we follow, which is very close to the law, is that a general expression against a country, for example, wouldn’t qualify as hate speech, but if you are promoting or advocating violence against a particular group based on their race or ethnicity, that would constitute hate speech. I accept that those are very nuanced judgments and that we are right on the borderline. As I mentioned, there are many videos relating to David Duke that go over the line and are removed. I would just emphasise that our teams are making highly principled decisions and debating these issues with a lot of intensity; we are not looking at these questions lightly. We are looking at the content of the videos and making decisions on whether or not they break our guidelines.

Chair

Have you watched this particular video?

Peter Barron

I have, yes.

Chair

Where in it is there actually something that justifies the title, “Jews admit organising white genocide”?

Peter Barron

I am not going to defend the content of the video; I found it abhorrent and offensive. However, the important question, which relates to wider issues of freedom of expression, is whether that content is illegal and whether it breaks our guidelines. Our policy and legal experts arrived at the conclusion that it didn’t. I think everyone in this room would agree that it was deeply distasteful.

Chair

But your own guidelines say that it is “not acceptable to post malicious, hateful comments about a group of people solely based on their race” or religion or so on. How on earth is the phrase, “Jews admit organising white genocide”, as well as being clearly false, not a statement that is a malicious or hateful comment about a group of people solely based on race, religion or the other protected characteristics that your own guidelines and community standards say are unacceptable?

Peter Barron

The test that our legal and policy experts are looking at is whether there is an incitement to violence against a particular identified group. I accept that these are borderline cases; we often see debate among our teams. The conclusion in this case was that it didn’t break our policy guidelines.

Chair

You don’t just say that, though. You don’t say you will only rule out stuff that is illegal or you will only remove stuff that somehow promotes violence. You have the phrase in your community standards, “it is not acceptable to post malicious, hateful comments about a group of people solely based on their race”, and yet you allow David Duke to upload an entire video that is all about malicious and hateful comments about Jewish people. How is that not a breach of your own guidelines?

Peter Barron

I am not sure that you have the guideline that is exactly the one that we use. I’ll see if I can pull it up. It is: “Our products are platforms for free expression. But we don’t support content that promotes or condones violence against individuals or groups based on race or ethnic origin, religion, disability, gender, age, nationality, veteran status, or sexual orientation/gender identity, or whose primary purpose is inciting hatred on the basis of these core characteristics. This can be a delicate balancing act, but if the primary purpose is to attack a protected group, the content crosses the line.”

Chair

And you don’t think that a video entitled “Jews admit organising white genocide” is one whose primary purpose is inciting hatred against a group of people?

Peter Barron

Ultimately, our lawyers, having looked at it and had a principled debate, ruled that that one did not break our guidelines.

Chair

Where are these lawyers and this group of people? Are they based in the UK? Are they based in the US? Where are they based?

Peter Barron

The lawyers and policy experts who were looking at this were based in the UK.

Chair

In the UK?

Peter Barron

Yes. Ultimately, these policy decisions are very often escalated right up to the top of YouTube and to the chief legal officer of Google and, on some occasions, to the chief executive. That happens on a relatively regular basis.

Chair

Do you agree with them?

Peter Barron

In this particular case? Of course I do.

Chair

So you think that those are not malicious or not hateful comments in the video?

Peter Barron

What I am saying is that it does not meet the test for removal under our guidelines. The point about all this is that we are in favour of free expression and we are in favour of access to information, but there has to be a balance. Let me quote what the Index on Censorship says about free expression. Free expression of course is an extremely important concept—

Chair

But I am not asking you about the law or about that; I am asking you about your own decisions based on your own community standards. You said a second ago that I was not necessarily quoting from the right community standards. This is from the YouTube website. It says in the section on “YouTube Help”, in the section on “Hate speech”, that “we don’t permit hate speech.” It specifically says that “we don’t permit hate speech. Hate speech refers to” and then it says what it refers to. Then it says that “it is…acceptable to criticise a nation state, but not acceptable to post malicious, hateful comments about a group of people solely based on their race.” So I am quoting at you your own community standards.

Peter Barron

I think if you go back one page, you will see the community standards, which are actually on the front page and which are what I quoted to you. We can come back to the Committee and clarify precisely the policy, but what I would say is that, as in the law, the emphasis is on promoting or condoning violence against a particular race or group.

Chair

Okay. I think it would be very helpful for the Committee to have a further explanation from you about why you think that that particular video does not violate your own community standards as well as, frankly, any standards of common decency, because I think most people would be appalled by that video and think that it goes against all standards of common decency in the United Kingdom. Let me turn to Facebook. On your website, in your community standards, you say: “Facebook removes hate speech, which includes content that directly attacks people based on their...race…religious affiliation…sexual orientation”—you then refer to other protected groups—“Organisations and people dedicated to promoting hatred against these protected groups are not allowed a presence on Facebook.” You still have on Facebook a couple of the pages that we raised with you. One is called “Ban Islam” and the other is called “The Truth About the Talmud”, both of which have some deeply offensive and racist things on them.

Simon Milner

Actually, all four of the pages that you notified to us are still on Facebook. May I briefly explain? Good afternoon, by the way. When it comes to reporting content on Facebook, most reports that we get are about individual pieces of content, rather than whole pages or whole groups. When it comes to the four pages that you notified to us, I am somewhat loth to talk about them in detail, because we don’t particularly want to give these pages a lot of publicity today, but if you want us to do that, we can. These pages, in and of themselves, do not violate, because we make it clear that you can criticise religions, but you cannot express hate against people because of their religion. “Ban Islam” is a page that is designed to criticise Islam as a religion; it is not expressly, in and of itself, designed to attack Muslims. That is also true of “The Truth About the Talmud” and the other two pages, regarding the Holocaust, that you also notified to us.

Chair

But you say as part of this: “Organisations and people dedicated to promoting hatred against these protected groups are not allowed a presence on Facebook.”

Simon Milner

That is correct. None of these pages are expressly run by hate organisations. There are hate organisations that we have identified that are not allowed a presence on Facebook. That is not true expressly of these pages—that they are run by hate organisations. But what I also can tell you is that we did look at the pages after you notified us of them. It was not possible to review every single piece of content on them in the time available, but there were certainly some that we spotted that had not been previously reported to us. We asked our expert team to look at them, and those posts were taken down. It is, relatively, a handful, and I am happy to talk a bit more about which ones those were.

Chair

Okay. I think we will come back to those in due course, because I want specifically to ask you some further questions about the concerns about your content monitoring more widely—concerns about it not working. You also have standards for things like dealing with sexualised images of children. You will be aware, obviously, of the BBC reports that they have reported to you 100 such images, that you have removed 18 of them and said that the remaining 82 did not breach your community standards. You then asked the BBC to send you the images that they were complaining about, according to the reports, and when they did so you then reported them to the police. That suggests that you thought these 82 images were somehow not worth you removing them from their pages, but they were worth reporting to the police when the BBC raised them with you. How come?

Simon Milner

Thank you for the opportunity to explain this. It needs a little bit of unpacking, so I hope you will bear with me while I do that. We welcome when a journalist or a safety organisation contacts us and says, “We think there’s something going wrong on your platform. We’re reporting something and your system is not working.” We welcome that because we know that we don’t always get it right. In this case, this particular journalist told us he had identified content which involved the sexualisation of children, but he was very clear that none of it was illegal. It was not illegal child sexual abuse material that he had found. He said it was, if you like, rather innocent pictures of children—probably, like me, you have those yourself, and you may have put them on Facebook, of the child at the swimming pool or on the beach—with horrible comments underneath them. So that is what he told us he had found. We asked him to send us the links, as we always do in situations like this, because then we can find the content very quickly. Also, especially if it is regarding child sexual abuse, as this Committee will know well, the law is very clear. If you send images of child sexual abuse, for whatever reason, that is unlawful. So we were very clear and said send us the links. Eventually this journalist sent us a document which actually contained screenshots. So he had been asked for links; we were sent screenshots. One of those screenshots—not 82, one—was an image of child sexual abuse, as determined by our expert team, and we were clearly under an obligation to let the authorities know about that, and it is now in their hands. It was also clear that the journalist had indeed identified something where our system was not working, and there was something going wrong. Through the other material he sent to us—and eventually he did send us the links—we were able to identify what the problem was. It was that the reviewers were not always being able to see the comments alongside the picture. So the picture was perfectly innocent and therefore he was getting a message back saying “This picture does not breach our terms”, but it was the connection between the picture and the comments, and something was not working in our review tool. We have now fixed that problem. All the content that he has told us about we have also addressed and reviewed, and it has come off Facebook.

Chair

So all the images that they sent to you have been taken down?

Simon Milner

All of the ones that that particular journalist has now told us about we have acted on. I do not know if it is all of the ones that he has talked about in his story; that is a matter I think we would have to ask him about, in terms of whether it is the full 100.

Chair

Some of the ones that they raised included groups, including closed groups, but groups that have titles like “Schoolgirls are hot in uniform” and really quite disturbing titles that seem to be linked to child abuse or inappropriate sexualised images of children, even if they do not have photos that might be publicly accessible. Do you have any process for assessing the names of groups or pages like that, and investigating?

Simon Milner

We do have some processes in that regard, regarding groups and pages. Clearly we do not publish the list of words that we particularly look out for. It could well be that even some words that will on the face of it look worrying—actually some of those pages are groups about highlighting this problem. But, yes, there are some things we do in the background to identify those.

Chair

But a series of these groups were reported to you some time ago and your system failed to respond and to do anything about them.

Simon Milner

In regard to the particular examples that this journalist identified to us, yes, something went wrong and we had to fix it.

Chair

So if I now did a search for these groups that exist, while we are in the course of this Committee, you are telling me that I will not find them.

Simon Milner

I cannot be absolutely authoritative on that, but I am happy to look into it for you and come back to the Committee. Equally, if you want to use your time now to do it, we could do that, but I cannot hand on heart tell you they are no longer there. This is a dynamic platform and people may try again to create bad content. That is why ensuring that our reporting tools are working and therefore learning from examples—learning from this particular issue that this journey has brought to our attention—should mean that things are much more effective. But I do not want to say to you, “The problem has gone away,” because when you have a platform as large as ours—we have got almost 1.9 billion people on Facebook—I am afraid that, much like we have 65 million people living in the UK, not all of them are good people. People are trying to do bad things, including to spread hate speech and foment disquiet. That is why we have to be ever-vigilant and not assume that we have fixed it—absolutely not. We always know that we can improve.

Chair

But do you have teams of people who would look themselves for that kind of abusive content or for groups and pages that have titles that frankly would disturb most parents across the country?

Simon Milner

We certainly do use technology to some extent in this area already. We deploy a technology called photo DNA, which you may be aware of, which enables us to scan every single image and video uploaded to Facebook against a database of known illegal images—child sexual abuse images. We have also just announced a collaboration with colleagues from Twitter, Google and Microsoft to develop a similar kind of system for the most extreme forms of terrorist content. Furthermore, our CEO, Mark Zuckerberg, wrote recently about what he sees as the potential for doing even more. So, in answer to your direct question: no, but that does not mean that we won’t have. But we actually think that technology here—because of the scale of our platforms—may be a more effective way of serving up likely infringing content to our reviewers. Human review will always be important, but it is about using technology to help you find that more effectively.

Chair

At the moment, are you confident that the kind of group where, for example, people can put up obscene sexualised comments alongside pictures of schoolchildren cannot take place on Facebook?

Simon Milner

I am confident that our systems for reviewing reports of such activity have been improved and that we are better able to address this problem, but I do not want to suggest that we have found a solution for a problem that occurs online—and, frankly, occurs in society—that absolutely prevents it from happening. I am afraid there are always bad people who will try to harm others, including children, and they will attempt to use the communication tools that we are all using, often for very positive things, for that purpose. We are absolutely determined to try to make it as hard as possible for those people to use our platform for that purpose.

Chair

How many people do you have working on this kind of thing?

Simon Milner

We have thousands of people who work in our community operations team, which is a global team with multiple languages who review millions of reports every week. They are based in a number of different locations, including Dublin—

Chair

How many people do you have in the UK?

Simon Milner

We don’t have anybody working in that team in the UK. The teams that do this kind of work in terms of reviewing reports have to have special kinds of access to the data of people on Facebook, which is not generally available to everybody who works on Facebook. They have to have that special access so they can look at absolutely everything, and therefore they are in dedicated centres of excellence in Dublin, California, Austin and Hyderabad in India.

Chair

So would complaints that are made in the UK—for example, complaints about groups coming from the UK—be looked at in Dublin, or would they be looked at in California?

Simon Milner

It would depend on the timing, what the language is and what the issue is about. Most reports that come from the UK, if they happen during our daytime, will be looked at by our team in Dublin, but it would depend on the particular issue or, frankly, the language—it may not be in English; it could need a language expert from somewhere else. So, it depends on the time of day and the topic of the report.

Chair

You think that you have now addressed the issue about context now being looked at. When did you sort that problem?

Simon Milner

I am afraid I cannot give you a definitive date for that, but it is certainly the case that this particular journalist’s work—especially once he finally gave us some links so we could actually understand what the problem was—we were on to straightaway in order to address it very quickly.

Chair

As of two days ago some of these, quite frankly, disturbing group names were still live on Facebook. We will send you some of those names separately and ask you to review them, because to be honest it just sounds like your response has been slow. Finally, as the result of journalistic investigation and having a Select Committee ready to ask you questions, you have looked at something. Given how long Facebook has been operating and how long there has been extreme concern about paedophilia and online child abuse, it is surprising Facebook has not looked at that before.

Simon Milner

With respect, I disagree. Jim Gamble, who is one of the country’s foremost experts on the problem of child sexual abuse and its behaviour online, was a very firm critic of Facebook six or seven years ago when he was running CEOP. If you talk to him now, he will tell you that Facebook has made enormous strides—not because of appearing in front of a Select Committee, as important as that is, or because of a journalist’s story. That does not mean we have fixed it completely, and therefore every moment like this is an opportunity to improve and to fix a problem—that is exactly what we do as a company.

Chair

We will move on to Twitter. We raised a series of images with you, most of which have been taken down. There is one that has not been taken down. Do you want to explain why?

Nick Pickles

Of the four accounts you brought to our attention, we suspended three when we were made aware of them. The one that remains—one of the things we look at is context, so the way that an image is communicated to another person as opposed to not to. There is a distinction between taking an image and targeting somebody who perhaps identifies with the protected category in that image, and posting it on its own. Many of the accounts you raised with us highlighted people who were posting images to other people who perhaps belonged to those protected categories. In this one case, the tweet that was reported was not directly sent to anybody else, and while it was highly offensive it did not breach our rules around hateful conduct.

Chair

So despite the fact that it includes a hashtag—“#DeportallMuslims”—with a very disturbing graphic cartoon that is clearly designed to alarm and distress, it does not breach your community standards?

Nick Pickles

In this context, we reviewed that particular tweet and image and found that it was not in breach of our hateful conduct policy. However, these are accounts where, in the example of “#killallMuslims” and hashtags like that, we do see offensive hashtags. “#BanIslam” was a hashtag that was trending after the Nice attack, but when you looked into it the overwhelming majority of the people using that hashtag were using it to condemn people who were using it to call out religion. While hashtags are offensive in themselves, often the conversation that happens around them is where people’s bigotry and intolerance is challenged. Yes, it is offensive but you will never get to a point where there is nothing on the internet that offends anybody, and nor should we seek to get to that point.

Chair

You have, “Examples of what we do not tolerate includes… behavior that incites fear about a protected group; repeated and/or or non-consensual slurs, epithets, racist and sexist tropes, or other content that degrades someone.” Are you really saying that the cartoon we sent to you is not racist, degrading or a repeated slurs?

Nick Pickles

As with colleagues, these decisions are very hard. There are lots of areas where it is very difficult to leave something up and it is offensive and sometimes upsetting, but yes, in this case we reviewed it and we did not find that it was in breach of our rules. Examples like this one are incredibly important to us because they are the examples we use to refine our policies. As a simple example, when Twitter first began we prohibited violent threats, so people would use words like “I wish” and “I hope” as a way of getting around that policy. This is not something that we say is finished; we are constantly thinking, “Is there behaviour that we would want to catch that is not being caught by a current policy?” Examples like this one go into that feedback process. At present it does not breach our terms, but I am not saying that it will never breach our terms.

Chair

How long does it normally take you to deal with reported complaints?

Nick Pickles

We want to get to every report as quickly as possible. One of the challenges about Twitter is that we see real-world events breaking on it. In the case of terrorist attacks, we will divert resources to deal with the reports coming in related to it, and that might mean that in other areas we are slower as a result. We take our resource and we prioritise accordingly. We prioritise reports of violent threats, for example. On your question about technology, one of the things we are trying to explore is how to use technology to better prioritise, so that we can be quicker. We have already made some changes internally on how we figure out whether two people have reported the same content, which several years ago we were not doing. We want to be faster, and we want to get to them quicker. We are using technology as well as people’s reports to do that.

Chair

But for example, yesterday the Committee sent you a series of individual tweets, most of which you responded to. I also, privately, on my normal Twitter account, reported various tweets to you that are still up from an account I found at random just by doing a little bit of searching for a few racist or violent hashtags. That account is still up, even though it includes one about Angela Merkel that says: “This German bitch needs a bullet in the head.” That seems to pretty much go against your community standards and is pretty vile. There is another one, similarly, with a violent threat against Gina Miller. There is another one with a threat against Sadiq Khan. There is another one with a series of vile, racist and violent attacks. Those are all still up, and I reported them yesterday. How long would you normally take to deal with those sorts of abusive tweets? I have not had anything back.

Nick Pickles

The challenge of communicating back with users is something we have acknowledged recently we are not doing a good enough job on. We don’t communicate enough with users when they report something. We don’t keep people updated enough, and we don’t communicate back enough when we do take action. I am sorry to hear that those reports have not been looked at. We would expect them to have been looked at, certainly by the end of today, particularly for violent threats. One of the things we are currently working on is how to use technology, as well as people, to identify those accounts for human review. For those kinds of violent threat that break our rules, we want to find a balance: as well as user reports, we want to proactively find those accounts for review, even if they are not reported. We have started rolling that technology out in recent weeks. That is a step change in how we deal with abuse—we are looking for it, and we will take action on content, even where it has not been reported by users.

Chair

Okay. This account is called @28sharwoddy. I hope that, if you have people at Twitter watching this sitting as it is streamed, they will do something about that and remove some of those abusive Tweets, even if they would not do so fast enough when an individual simply reports them through the normal processes. There is a broader point here. I have been going through the individual examples. To be honest, I find none of your three responses particularly convincing. We understand the challenges you face. This is all moving very fast, and technology changes very fast, but you all have millions of users in the United Kingdom, and you make billions of pounds from those users. You all have a terrible reputation among users for dealing swiftly with problems in content, even against your own community standards. Surely when you manage to have such a good reputation with advertisers for targeting content and doing all kinds of sophisticated things with your platforms, you should be able to do a better job in order to keep your users safe online and deal with this kind of hate speech.

Simon Milner

With respect, I disagree. I spend a lot of my time in the UK, and indeed in other countries, listening to people about their experiences on Facebook, and I particularly encourage people, even if they have never had a problem with Facebook, which many people have not—unfortunately that is not true of many politicians—to report and to try our reporting tools. I get some very positive feedback from people about their experience of doing that. That is not to suggest that we get it right every time. When you are dealing with millions of reports in multiple languages about a range of issues every week, you cannot possibly get it right every time. We would love to be able to get it right every time and to make the correct judgments each time. We want to improve, but to suggest that we are in some way negligent or not caring about this issue is simply not true. We deploy a lot of time, effort and resource to try to improve what we do in this area.

Peter Barron

We are constantly making improvements in this area. For example, I would point to the German tests that have been going on today, which were just announced. That was for illegal material that was flagged to our systems, asking us to take it down as quickly as possible. I don’t know if you have seen the results yet; they were just announced in the last few hours. They have shown significant improvements in terms of the material that we have been able to remove—around 90%—so we take these issues very seriously. As Simon says, we are never going to get to 100%. The platforms are very big and there will always be elements of unpleasant material on them. In terms of the way we have approached these issues over the years, on a range of different issues—for example, going back a few years, the work we have done since 2006 on child sexual abuse imagery—we have significantly improved our approach and our ability to remove that content. Some 98% were reviewed within 24 hours, which is a pretty high bar. It is something that we have made a lot of progress on.

Nick Pickles

No. I think this is an issue where Twitter, as a company, has acknowledged that, in the past, we haven’t done enough. It is something that, from the top of our company, we have said is a priority. We are now in a renewed stage of not waiting months to change things; we are changing our product every week right now. We announced new technology in February to stop people repeatedly creating accounts. We have said we are going to proactively look for accounts that are actually posting this kind of material and behaving in a certain way and limit the number of people who can see those accounts. We have also announced new tools to allow people to say, for example, to never show them a tweet from someone who has never changed their profile picture or someone who hasn’t verified their phone number. We are now changing our product every day, which is partly a recognition that we haven’t done enough. I think there is a broader question that. For all the technology we can deploy, and however good our policies and systems are—no system is ever 100% accurate—a lot of the issues that we face are based on real world prejudice and ignorance. By removing content, our platforms are not going to remove that ignorance from society. That is one of the challenging things we have to face, because there will be things on our platforms that we personally may find offensive, but if someone is expressing an offensive view, the best challenge to that is for people to challenge those views. That is how we get progress. It is painful; it is sometimes upsetting. We need to do more as a company to protect it, but the essence is that you cannot solve social issues with technology.

Stuart C. McDonald

May I just ask you a little bit about some recent high-profile examples in which individuals have been targeted? We are talking about, perhaps, musicians, singers or political figures. Diane Abbott was an example. On Sunday—I don’t know if you are aware of the story in the Sunday Herald in Scotland—there was an article about the catalogue of horrific online abuse against political leaders such as Nicola Sturgeon, Ruth Davidson and Kezia Dugdale, which included death threats, threats of violence and extremely offensive and sexualised insults. Were you aware of that particular story? Twitter was asked to comment on it. Did you get to see that story that appeared on Sunday about those Scottish political figures?

Nick Pickles

I don’t know if it was Sunday, but I saw something about Scottish political figures. I thought that was today, actually.

Stuart C. McDonald

So you wouldn’t be able to say if anything has been done about these tweets since then? There were 150 tweets highlighted.

Nick Pickles

I don’t know the details of it, no.

Stuart C. McDonald

Sure. More generally, then, are there proactive steps that Twitter can take when it is clear that an individual has been targeted? What is the company able to do if it is becoming obvious that individuals, such as Diane Abbott, or musicians or political leaders are being targeted pretty much every day?

Nick Pickles

This is one of the areas in which we think technology has a role to play. We have already deployed the very first steps of using technology in these cases. There are two things. First, it is about identifying offenders, and identifying people who perhaps have several accounts. We have built technology to spot people who may use several accounts to send abuse to one person. We want to remove those people from our platform. As I say, we have tested and are rolling out technology to limit the number of people who see tweets. You can still tweet, but it might not show up in the politician or singer’s timeline. I think we recognise that, in those situations, when you are having a lot of incoming messages, we need to find a better way of not having the burden being fully on the victim. We are experimenting with technology that way, and also to help victims to control what they see, which might be by prompting someone to ask whether they would like to turn on a setting. We are looking at technology to both catch the offenders and to help victims in those circumstances. Some of this comes down to very contentious political issues. We have seen more seismic political debates in the past 18 months than many people will see in a lifetime, from Scottish independence through to Brexit and the refugee crisis. Some of those comments are going to be offensive and very challenging but they will not break our guidelines. That is where we need to get better at spotting the people who do.

Stuart C. McDonald

In relation to your guidelines, I was interested in what you said about the difference between a tweet that was directed or sent to somebody and one that was not. I will pick out one of the milder examples of a tweet here. “If—I will not name the politician—was in front of me. I’d kick her something head in and probably really enjoy myself doing so.” That is not tweeted at anybody, and it does not use a handle or a hashtag or anything. Does that breach your guidelines, because it is sent out without reference to a handle or a hashtag?

Nick Pickles

As I said, it is one of the things we take into account when we look at an account. It is not the predominant factor. We will look at a variety of things, as I say. We might look at: does someone have several accounts? We might look at: has that person sent one tweet or have they sent a series of tweets? One thing we have introduced is a time-out, to say, “You just can’t tweet for a certain time period.” That case you described sounds more severe but it is one where we can see a situation escalating and say, “Can we intervene early, stop someone tweeting and stop the situation escalating to that kind of level?” It is still very important that users report content to us. We review content every day. Where it breaks our rules, we will take action.

Stuart C. McDonald

If some of these tweets do not in themselves breach your guidelines or policies but the person who is tweeting is breaking the law of Scotland, the United Kingdom or whatever, does that affect how you then deal with these individuals? Or do you just take an approach that that is a matter between them and the police and it is not for you to interfere with that?

Nick Pickles

We do work with the police. For example, if the police give us court orders. Some of the work we have been doing that we can perhaps talk about later is with the Crown Prosecution Service to talk about how restraining orders can be better used online. The Government says in its own written evidence to this Committee that the vast majority of what we are talking about is not criminal. That is something that we see. Where there are allegations of illegality, it is not appropriate for companies to be making decisions, I think. In the case of criminal allegations, it is correct to follow due process and, where we receive due process that the law has been broken, we will take appropriate action. People are also getting better at including in court orders the requirement to remove accounts upon conviction. I think that is something that should be encouraged.

Stuart C. McDonald

Are you saying there would not be any cases where, for example, a tweet might breach the law of Scotland or the UK but still not fall foul of your community guidelines? Or would you think that is inconceivable?

Nick Pickles

It is not helpful to get into hypotheticals and try to think of situations where that could happen. In my time at Twitter, I can’t think of a case where I have seen something that has been the result of a criminal prosecution that has not breached some of our terms of service.

Mr Umunna

May I come back to Mr Barron about one of the cases we have been looking at? Can I ask you a basic question? Google is a for-profit company.

Peter Barron

Of course.

Mr Umunna

What was Google’s operating profit in the last financial year, 2016?

Peter Barron

Globally, around $20 billion.

Mr Umunna

$20 billion? I have $30.4 billion here—figures that the House of Commons Library has produced.

Peter Barron

We can debate the detail.

Mr Umunna

Billions then. How much does whoever posts a video on your site make per view from adverts that are posted alongside YouTube videos?

Peter Barron

Per view would be a very tiny amount of money.

Mr Umunna

I think it is $7.6 for every thousand views. Is that right? Does that sound in the ballpark?

Peter Barron

It is a reasonable figure, yes.

Mr Umunna

Supporters of ISIS and neo-Nazi groups have been posting videos on your platform and taking up the option of making money from ads that appear alongside their videos, which makes them and you money. That has happened, hasn’t it?

Peter Barron

It is very, very important to point out that terrorist organisations or their agents are not allowed to have YouTube channels and, therefore, not allowed to make money. Wherever we see that a channel—

Mr Umunna

I was quite careful then so let me clarify. I’m pretty sure I said supporters of ISIS and neo-Nazi groups.

Peter Barron

We wouldn’t know what the background of an individual was, but we do not allow hate speech or terrorist content to be monetised. Whenever that is drawn to attention it is, of course, shut down.

Mr Umunna

But I asked you a factual question, and you will be aware of the important investigation that The Times has done on this. Going back to my question, supporters of ISIS and neo-Nazi groups have posted videos on your platform alongside ads that have made them and you money. Is that correct?

Peter Barron

Let me go into detail about The Times investigation, which is clearly what you are referring to. It is a relatively complex picture. “Big brands fund fear” was clearly a compelling headline, but if you dig into the detail of the videos involved in that case, it is a much more complicated picture. Some of the videos were against our guidelines, both on the content and monetisation side, and were removed. Others broke our advertising guidelines but not our content guidelines. Our advertising guidelines are more restrictive than our content guidelines and in some cases, videos that were drawn to our attention by The Times didn’t break our guidelines—

Mr Umunna

Sorry to interrupt you, but I will be cut off by the Chair unless I get to the point. To go back to the question, I have seen the videos that The Times refers to, as I am sure many others have, and there is no getting away from the fact that supporters of ISIS and various neo-Nazi groups have posted video content on your platform with ads going alongside it that have made you and them money. All I am asking for from you is a straightforward answer. That has happened, has it not?

Peter Barron

Let me give you an absolutely straightforward answer. The video on the cover of The Times underneath the headline in that particular case made no money for the uploader. The monetisation on that related to the music on the video.

Mr Umunna

Hang on just a moment. Money was being made from videos peddling hate on your platform.

Peter Barron

Very small amounts of money. I don’t want to mislead you but in terms of the amounts of money made from the videos in question, we are probably talking about around ten—

Mr Umunna

But you don’t dispute the fact that you and these hate peddlers have made money out of ads appearing alongside those videos? That has happened on your platform.

Peter Barron

I would—

Mr Umunna

I am asking for a straightforward answer.

Peter Barron

I don’t dispute that, on occasions, of course that will happen. I should say that advertisers have a large degree of granularity and control. They can choose not to appear next to political speech—

Mr Umunna

The point is that you are providing a platform that has acted as a money-making machine for the peddlers of hate and extremism, for supporters of ISIS and supporters of these neo-Nazi groups. That is happening on your platform and the way in which you are prevaricating and dancing around this factual matrix is disturbing. If I am honest with you, Mr Barron, all you need to do is say, “Yes, that has happened and this is what we are doing.”

Peter Barron

We have no interest in making money from that.

Mr Umunna

But you have.

Peter Barron

It has happened. We work very hard to make sure that it does not happen, and we work with advertisers to give them more transparency and control so that they don’t appear next to political content. It is worth pointing out that some of the videos that you are referring to were not ones that would break our guidelines. There is a whole area of concern about these kinds of videos—Nasheed videos, for example, are Islamic songs that may be referring to history, battles and so on. These videos may be distasteful to some and may be talking in broad terms about jihad, but they would not break our content policies. It is impossible to say, of course, who uploaded them. Where a terror or illegal organisation attempts to have a YouTube channel and to make money from that, it would be shut down and drawn to our attention.

Mr Umunna

My final point is this: the bottom line is that you are a for-profit company. Your operating profit in 2016 was $30.4 billion. There are not many business activities where somebody, openly, would come and give evidence to this Committee and have to admit, no matter how many times they dance around, that they are making money, and people who use their platform are making money, out of hate. That is happening on your platform.

Peter Barron

We never want to make money out of hate.

Mr Umunna

And I would suggest to you, Mr Barron, that you as an outfit are not working nearly hard enough to deal with this.

Peter Barron

We are working very hard in this area and we are investing hundreds of millions of dollars in the technology and teams aimed at addressing these issues.

Mr Winnick

Do you feel any shame at all about the things that my colleagues have been referring to and what has been distributed on Facebook and Twitter? Do you feel any shame whatsoever?

Simon Milner

Would you like me to take this, given that my colleagues have had something to say? Perhaps I could just talk about the two—

Mr Winnick

No, just answer the question. Presumably the answer is no. You feel no shame, do you, Mr Milner, Mr Barron, Mr Pickles? Just answer the question without going into a rant about this, that and the other.

Simon Milner

With respect, I have maintained—

Mr Winnick

You always preface your comments with “With respect”, but it’s a simple question. Do you feel any shame at all about some of what is seen here?

Simon Milner

Mr Winnick, I don’t think it’s a simple question to say, “Do you have no shame?” I feel very responsible, as do my colleagues, for the safety of the 1.9 billion people using our service, and I am very proud of the work that we do.

Mr Winnick

Proud of the work? I see. Well, I am looking at some of the items that have been sent to us. Mainly, it’s Twitter. And what I see before me is a group of Muslims engaged in rape, and the person who is complaining that it’s his daughter who is the victim is told he is a “Racist bigot!” Then I see other items, equally obnoxious—for example, on Facebook, Mr Barron.

Simon Milner

That’s me.

Mr Winnick

Sorry, Mr Milner—apologies. On Facebook, I see a Muslim at prayer, and while he is praying there is a dog behind him—right behind him. No shame—no apology?

Nick Pickles

With respect to the comment that you have just referenced on Twitter, we have removed that, because it broke our rules. What is incredibly important here is to remember that we have people around the world who work very, very hard to address these concerns. This isn’t just us, sitting in London—

Mr Winnick

So hard that it appears before us now.

Nick Pickles

It has been removed, because it broke our rules. It is in your dossier, because your dossier was produced before we—

Mr Winnick

But it appeared in the first place, didn’t it?

Nick Pickles

Let us be absolutely clear: we are never going to get to a point where internet companies pre-moderate content for the 400 hours of YouTube going up every day and for the 500 million tweets that go up every day. If you want pre-moderation of internet platforms, there may well be no internet platforms. I think we need to be very, very clear about how we discuss this, because there is a scale challenge. The positive benefits that our platforms bring and technology brings—yes, it comes with serious challenges. Yes, it brings out some of the worst in society and it brings to light things that we would all rather did not happen. But the idea that you can pre-emptively detect things and then remove them before they have been posted—we are never going to get to that point, and I think we need to be honest about that.

Mr Winnick

Listening to your response to the question from my colleague, Mr Umunna, when it came to the amount of money made, the billions of dollars, the thought that came into my mind was that it’s a form of commercial prostitution that you are engaged in. I think that is a good and apt description. For example, Mr Barron, the way in which you did your utmost to evade the question put to you by the Chair about that obnoxious entry in relation to your social media organisation—it was a rather pathetic answer. I just wonder whether, on reflection, you don’t feel you could have given a far better, more forthright response to the Chair.

Peter Barron

This is, of course, uncomfortable territory in the area of free expression; there is no question about that. The point I was making to the Chair was that these are principled decisions and sometimes extremely borderline decisions, but as an organisation we have a bias towards free expression—we are in favour of free expression—and we are in favour of access to information. As Nick says, there is an issue about how these kinds of beliefs and behaviours exist in wider society, and as an organisation we think that the best way of addressing these issues is to shine a light upon them and allow them to be debated and held up to scrutiny and ridicule. Of course these are difficult decisions but—

Mr Winnick

What you are saying, in effect, is that it should be put on as it has been and subject to ridicule, but it should be allowed to be there in the first place.

Peter Barron

If content is illegal, of course it should be removed. As in society, the same thing applies online. Illegal content is not allowed and will be removed; content that is offensive and extremely unappealing but nevertheless legal, we judge should stay on the platform. Of course, counter-speech—this is a very important area—is the best antidote to these kinds of beliefs and thoughts. We invest a huge amount of effort in working with organisations such as the Home Office and the police in promoting counter-speech. We have a project called the Redirect Method, which looks at promoting content that already exists on YouTube and that puts across alternatives to some of the very unpleasant thoughts that are out there. We work very effectively with influencers—with creators—on YouTube who are very trusted and have a lot of influence among their fans, to promote more empathy and tolerance across the platform.

Mr Winnick

Would it be more correct to say that this deeply obnoxious and racist propaganda and incitement to race hatred—there is no other way to describe it—goes on, and it is only under public pressure that you take any action, but the internal policing is minimal?

Simon Milner

That is simply not true. As I mentioned earlier when the Chair asked me, we have thousands of people working in our teams that are reviewing reports from our users, and we get millions of reports every week. In fact, the last time I was here before this Committee—albeit with a different Chair—I said that you would be very welcome to come and visit us in Dublin. Come to our international headquarters, meet those teams, hear from them about the work they do and see the expertise they have. You could also visit Google and Twitter at the same time. You would really learn from the horse’s mouth, not just from the spokespeople here in London, about the work that those teams do. We would be very happy to welcome you. May I give you an example of counter-speech? One of the pages you identified for us—a Holocaust denial page—only has nine followers. It was posted in 2013, but the second comment on the page is, “I have relatives who were there and saw everything you thick”—I won’t use the expletive—“It happened whether you like it or not. Go away and spread your hatred somewhere else where there are small-minded people like you.” You see that all the time on Facebook—people who want to try and divide us, and other people who say, “No, that is simply not true. I have friends from that community. I have family who were there at the Holocaust. You are talking nonsense.” That is much of what you see in response to this material that does not breach our hate standards because it is trying to suggest an alternative view of history—you actually get lots of content from people saying it is nonsense.

Mr Winnick

Mr Milner, Holocaust denial is not a matter of debate; it happened. Therefore to say, in effect, “Well, it leads to a debate. Did it or did it not happen? Were millions of people murdered because of their racial origin?”—that is a not a subject of some sort of debate; it is a fact. If you give a platform to racists, and it leads to that type of debate that you are taking, it minimises one of the worst atrocities that has ever taken place in Europe. I do not think that is a good example. May I just say this in conclusion? I won’t put it in the form of a question. I would be ashamed—absolutely ashamed—to earn my money in the way in which you three do.

James Berry

Mr Pickles, using the phrase “Filthy Jew bitch” face to face would be unacceptable and in most circumstances would be unlawful, wouldn’t it?

Nick Pickles

I cannot speak for the quality of the law. That is a question for a police officer.

James Berry

So you have no idea whether that would be unlawful in some circumstances.

Nick Pickles

I imagine that the answer would be that it depends on context.

James Berry

So in some circumstances it would be unlawful.

Nick Pickles

Perhaps.

James Berry

Okay. Using that phrase on Twitter would be equally unacceptable, wouldn’t it?

Nick Pickles

Again, it depends on context. You could be condemning it.

James Berry

In the context where you are not reporting speech or condemning it, using that phrase would be completely unacceptable, wouldn’t it?

Nick Pickles

I expect so.

James Berry

And in some circumstances unlawful.

Nick Pickles

Again, that is not something that I should speak to.

James Berry

Can you confirm that you are aware that this Committee has raised concerns about that phrase being used on your platform? You are aware of that, yes? So you would be disappointed if that phrase was still being used on your platform, wouldn’t you?

Nick Pickles

Yes, I would.

James Berry

So are you aware that that phrase, or variants of it, have been appearing on your platform since 2014 and are still there today?

Nick Pickles

Specifically with regard to this case, yes, we actioned a large number of accounts that had used that hashtag abusively. We have 500 million tweets every day and I am certainly not sat in my office looking through them for particular hashtags. I would say that where people have reported those accounts, they have been taken down. I believe that there were criminal prosecutions connected with it. Also, if that term was targeted at somebody and used in an abusive manner, as it often was, it is likely that those accounts breach our terms of service. There may still be tweets on our platform that use that hashtag that do not break our terms of service—I would have to check—but my expectation is that the ones that were posted abusively were actioned and many accounts were removed.

James Berry

I did a bit of searching about an hour and a half ago without the help of any technology that you might have or any staff, and from 11 March there is a tweet, “Filthy kike bitch. Burn in hell, you fake Jew.” That would breach your community terms?

Nick Pickles

I expect so.

James Berry

From 17 October last year—it has been there for a while—“high speed picture of filthy Jew getting bitch-slapped”. That would breach your community terms of service, wouldn’t it?

Nick Pickles

I would refer it to the teams, but absolutely it would, I expect.

James Berry

Another one—I will not mention the MP’s name, but I think you know who I mean—“filthy Jew bitch—MP’s name—attacks free speech. Once again demands a nationalist rally be shut down.” That is from 2 February 2015. That would breach your community terms of unacceptable abuse, wouldn’t it?

Nick Pickles

I expect so.

James Berry

And one from 15 November 2014, referring to the same MP’s name, “Filthy Jew bitch. Her dirty games caused labour councillor to quit politics”. That would also breach your community terms.

Nick Pickles

I would refer to the teams, but I expect so.

James Berry

Why is it then, just to take the last tweet, that that tweet is still there from November 2014?

Nick Pickles

I cannot speak to individual accounts and individual tweets. I would certainly be happy for my colleagues to review those accounts and write to the Committee to update them on what happened with them.

James Berry

It seems strange that, when that particular phrase has been raised—I have raised it on the Floor of the House—at least once in a report of this Select Committee that you said you were aware of, that it took two seconds to find multiple tweets using that phrase. Is there any reason that Twitter, with all your resources, could not have removed all those tweets, almost instantaneously?

Nick Pickles

I can ask my colleagues to review why that did not happen. I know that we have actioned numerous accounts relating to that specific case of abuse. The overwhelming likelihood is that the examples that you have raised most likely do breach our terms of use, and I will ask colleagues to review them.

James Berry

But it is completely unacceptable that they are still on Twitter’s platform, certainly from 2014, if not from a few days ago.

Nick Pickles

One of the critical questions for us is that we are currently trying to experiment with how we use technology proactively as opposed to waiting for users to report things. I cannot speak to which tweets were reported, if they have been reported, but I am happy to investigate them and update the Committee afterwards.

James Berry

I found these and many other tweets in a matter of seconds with no resources. How many staff does Twitter have looking for this kind of material, performing even the very simple searches that I just did?

Nick Pickles

One of the things to highlight is that we don’t proactively look for tweets. We don’t have people sat searching Twitter. We are currently experimenting and there is huge potential in using technology to do that, but we are in the very early days of delivering it. Since that first tweet you cited from 2014, we will have seen hundreds of billions of tweets being posted, so there is a not insignificant challenge of scale. We absolutely think technology has a part to play, but we rely on user reports as the primary means at present. In the long run we hope that technology can do a lot of that work, but artificial intelligence and machine learning is not a switch you can turn on and suddenly let it run. The technology is very challenging.

James Berry

Not searching proactively is a policy decision by Twitter, isn’t it?

Nick Pickles

Some of it is a question of limitation, but yes. It is the same with colleagues—the “notice and take down” model underpins most internet companies, not just our own.

James Berry

Right, so it is a policy decision, but the resources that would be required to do what I did are minimal. We know that certain words and phrases are closely associated with racism. It is very easy to perform a quick search and then to manually filter out the ones that are reported speech or are condemning the use of those kinds of words. It didn’t take me long.

Nick Pickles

No, and you have used one example. Sadly, there are examples around the world every day—there would be more than a couple. We could have staff, first, developing technology to help users to find the content and secondly, reviewing reports. Or, with 500 million tweets posted every day with different hashtags, staff could say, “I agree that this one hashtag is very focused,” but looking at every hashtag on Twitter in that way is not a resource-scalable solution. I don’t think it is for any business, so we need to figure how to use technology and our user reports to better action our terms of service and our rules. That is the challenge for us.

James Berry

Do you think it is possible that you have allowed the Twitter membership to grow too big for you to be able to do the level of searching that is required to make sure that that kind of phrase is not being used or is removed promptly?

Nick Pickles

That is why we are looking to invest in technology right now. We are looking to build things that we can use, not in months and years, but in weeks and days, that will take the burden off users when they are on the receiving end of the situations that your colleague, Mr McDonald, outlined, and that limits the ability of people to send tweets to other users and finds things that may end up with users being removed from our platform. That technology is being developed and deployed right now. Are we doing it fast enough? We would like to do it faster, but this technology—machine learning and natural language processing—is some of the hardest technology in the internet space, and it is not something that we are going to crack overnight.

James Berry

The Met police have dozens of staff who search for extremist and offensive content, and they report it to you for breaching your community terms of use. To your credit, you remove most of it, so it is possible to do that. My last question to you is this: who should be paying for identifying that material? Should it be my constituents through taxation or the police precept, or should it be Twitter?

Nick Pickles

In the context of CTIRU, we have a great dialogue and discuss various issues with them. At this point, the vast majority of accounts that we receive from CTIRU have already been suspended by us by the time that we open their reports. To pick up something that Simon said earlier, to take the case of terrorist accounts, the majority of accounts that Twitter removes for promoting terrorism are detected and removed by technological means. This is something where industry is acting. We are collaborating with different companies and are using technology to find content so that when the police come to us and say, “We think we found something,” we can say it is already suspended. All the accounts that CTIRU have submitted to us in recent reports have been suspended prior to them submitting their reports.

James Berry

To return to my question, you, for policy reasons, do not search for things proactively at the moment. Most people in the country think that you should, and our police are doing that for you. Do you think that we should be paying for that, or do you think that you should be paying for it as the company that profits from providing the platform on which this kind of behaviour can take place?

Nick Pickles

There is a really important question about how the internet fits into society, because the same argument is true of cars. Should car manufacturers pay for transport policing? The same is true of every major technological development that has come along. The internet is the latest one. Is the internet part of people’s everyday lives? Yes. Does it, therefore, have a policing element to it? Yes. Again, I refer you to the Government’s own written evidence to this Committee, which says that the vast majority of things that happen online, even where they are abusive, are not criminal.

James Berry

The difference there, of course, is that car manufacturers meet a lot of health and safety regulations put on them by the Government but you do not have any similar regulations that you have to meet. You are effectively unregulated in this respect. I return to my question. Our taxpayers’ money is funding police to police the content on your platform. Should taxpayers be paying for that or should you?

Nick Pickles

One thing you have heard us all say is that we all have standards. We can disagree about whether those standards go far enough but we all have standards. We all have teams who are looking at removing content. We all action content that does breach our terms and conditions. I think the industry is actually stepping up and saying we can do more. We are looking to be more proactive and use technology better. I don’t think this is a simple either/or. It is industry saying, “We do have a responsibility here and we are going to continue to look at ways where we can better fulfil that responsibility.” But we will never be done. We will never say, “Safety is finished; we can stop thinking about it.” This challenge continuously evolves and behaviour changes so we will keep working hard to keep our platform safe.

James Berry

But if the police submitted a bill to you for the 90% of material that you do take down—and therefore agree breaches your community use standards—would you be willing to pay it?

Nick Pickles

As I said, the vast majority of accounts that CTIRU report to us right now are already suspended.

James Berry

I am asking about the ones that you do take down; the ones where you do agree that police have identified something that needs to be taken down. If they submitted a bill to you for the work involved in doing something that a lot of people would say that you should be doing yourselves, would you be willing to pay it?

Nick Pickles

That is a conversation we can have with the police.

Chair

We would ask all of you the same broader point that Mr Berry is raising about proactive identification of problems, at least in the most extreme cases, whether that be terrorist activity or proscribed organisations, online child abuse, or where a known trope or meme that is being used often to target individuals. It is baffling to us that you do not seem to have any proactive activity in these extreme cases, even if you are not doing a broad sweep of everything that comes in. Where you have these extreme cases, where everybody accepts that this is about illegal or really problematic content that clearly goes against your own community standards, we don’t understand why you don’t have any proactive activity.

Simon Milner

Chair, I think I explained earlier that we do. When it comes to child sexual abuse, absolutely we do. When it comes to terrorism, we do. We announced the EU Internet Forum in December—this collaboration among our companies to share information that would enable us to spot attempts by global terrorist organisations to distribute new extremist propaganda. Wherever it appears first, we are all able effectively to nuke it before it gets purchase. What we also do when that happens or when we come across terrorist content, we will fan out from that account. We will look at the groups that person is in. We can often take down many, many more accounts as a result of a new piece of intelligence. There absolutely is proactive work, and by human beings. What we also want to do—and we have talked about it—is to explore some of this really cutting-edge new technology around artificial intelligence. How can we surface? How can we improve the funnel, if you like? The reporting we get from the public is like a funnel. It is not all wonderful; a lot of it is noise. With any kind of reporting, you want your expert people to be looking only at reports where there is a very high chance that it is infringing content. Otherwise it is a waste of their time. If we can make that funnel better using technology and still have human review, that could be a much more effective and, indeed, proactive way to do it. We are already doing proactive work but we want to do more.

Chair

So Facebook does proactive searching if you have particular issues on child abuse or you have a report that comes to you that raises problems? You will then do proactive searches linked to that?

Simon Milner

We will fan out from that, particularly in these extreme areas where, as a society, we are most concerned such as child abuse, terrorist activity and hate organisations. Absolutely, we fan out.

Chair

Let’s clarify this for each of the three of you. On child abuse issues, do you do the proactive searching for frequent terms used around child abuse?

Nick Pickles

I will speak for myself. We are all members of the Internet Watch Foundation, which offers a variety of different products for different types of businesses. They offer a URL list, which is a website list that is predominantly used by ISPs. One of the technologies that we have been using is the hash database, which is where known images on one platform—

Chair

Sure. I understand about the—

Nick Pickles

That is all proactive.

Chair

Obviously that is easy and you can do it on a technological basis. What I am interested in is whether there are particular phrases that are often used, or if you actually just do a search from time to time to see what is coming up—perhaps around child abuse or things that have been raised—which is not stuff that would be automatically caught by the Internet Watch Foundation work.

Nick Pickles

They actually offer a search list as well, for that reason. As part of their suite of products, one of the things that they offer members is a keyword list of that variety, which—speaking for us—we have certainly used.

Chair

And that is done continually, is it?

Nick Pickles

Yes.

Simon Milner

I can’t be certain on that—I would like to write to the Committee. What I can tell you is that we have a very close working relationship with CEOP and other forces around the world that play a similar role to CEOP. We employ someone who used to work at CEOP in order to be that liaison point with them and pick up what is going on—are there persons of interest, is there a new activity that they are seeing that might be happening on Facebook that we should be looking out for? It is not just about words; it is also about understanding what is going on in the real world and perhaps on other platforms, to make sure that we are looking out for new kinds of behaviour by these individuals, who after all are often very skilled at trying to evade detection, whether by the police or by our experts.

Peter Barron

As colleagues have mentioned, we use technology where it can be effective. We are able to identify spam and abusive accounts and automatically get rid of those. On hashing technology, which we have been discussing in some detail, we work with the Internet Watch Foundation, in particular for the hashing technology around child sexual abuse imagery. As we have been discussing and as the European Commissioner announced last week, we are embarking on the shared hashing database around terrorist content.

Chair

But you don’t do things like look up National Action, from the discussion we had at the beginning of this session.

Peter Barron

No, we don’t. Again, this is the principle of notice and take down. If we go back to the principles that were put in place with the e-commerce directive, the intention for how the web operates is that platforms should take down—

Chair

Sure. Other Committee members want to ask questions; I just want to understand the factual position. Do you do proactive searches and investigations on child abuse, on terrorism and on particular targeted hate crime phrases or attacks like “#FilthyJewBitch”, which Mr Berry referred to?

Peter Barron

I wanted to explain what the principles of the way we operate are, because the reason that—

Chair

I just want a yes or no, because we have a lot to get through. Do you do proactive work on those three issues?

Peter Barron

As I explained, we use technology proactively where it is effective. We do not proactively search for—

Chair

So you don’t proactively search for any of those three things.

Peter Barron

Because of the underpinning of the e-commerce directive.

Chair

Sorry, the e-commerce directive stops you from doing those proactive searches?

Peter Barron

The principle of notice and take down is extremely important, because the way that the rules were designed was that as long as the platforms have a responsibility to take down when given notice, they do not have a liability for all the stuff that might appear on them. Of course, the point of this is to allow people to share content, to share creativity. The web would not have grown—it would not have been possible for Twitter, Facebook and YouTube to—

Chair

So you, as a company that provides a platform for a huge number of people and makes multi-billion pound profits from it, do no proactive work at all on clearly illegal content, including terrorism and online child abuse.

Peter Barron

We are involved in a huge—

Chair

A yes or no answer, please.

Peter Barron

We don’t proactively search for this content.

Chair

Mr Pickles, do you have anything to add about any proactive work that you do?

Nick Pickles

I think I have covered some of the new proactive work that we are doing in the abuse space, but we also use PhotoDNA, as Simon says, as part of addressing child sexual exploitation, and we are part of the proactive hash-sharing programme on terrorist images.

Chair

So you do proactive work on terrorist—

Nick Pickles

And on child exploitation and abuse.

Chair

Okay. We would like further clarity from Facebook and Twitter on the precise proactive work that you do. While we have been sitting here, I have done some obvious searches that you might do on Facebook on phrases around schoolgirls. That has come up with a series of pages that I think it would be worth Facebook looking into, because it does not look as though the proactive things that you would expect are necessarily filtering things out. Similarly, on Twitter, it would be useful to know precisely what proactive work you do.

Naz Shah

As it is, if I want to open a Google or Twitter account, I don’t even have to be real, do I? I can use a pseudo-name and a pseudo-everything. Yet when I want a verification on my MP Twitter account, I have to go through lots of hoops to get it. It feels to me that you guys—your companies—have created a second world that is anonymous and you are not doing anything to stop that happening. First, you are absolutely failing to deal with what you have already got on there. Secondly, what are any of you doing to stop new accounts? I will give an example from Twitter. TellMAMA complained about @DowHeater for four years. You finally took it down and that person now has four more accounts. What is the point of that?

Nick Pickles

Our policies explicitly prohibit reopening a new account once you are suspended. This is one of the challenges that we live with: when we suspend an account, someone will create a new account, then we will suspend that one and we get into this cycle. As Simon mentioned, some of these people are technically determined, so they will try to find ways to get around the measures that we put in place, whether they be phone or email verification.

Naz Shah

But I had to go through hoops and loops to get my verification. Why don’t they have to apply the same standards to get a YouTube video or a Twitter, Google or Facebook account, and yet we have to do that to get verified?

Nick Pickles

But one of the values of having a system of verification is that users can see that this person’s identity is real: they have a blue badge, they have submitted identification and we know who they are. That is why we are also now saying to people, “If you don’t want to see tweets from people who haven’t verified that their phone number is real, you can tick a box and you won’t see them.” It is why we are also saying, “If you don’t want to see tweets from people who have not changed their profile picture, you can now get rid of them.” That is in the process of going live as we speak, so there is technology there.

Naz Shah

But isn’t that a huge contradiction? On the one hand, you are saying, “Let’s verify”, but on the other you are letting numerous accounts just come online every single day of the week. Every hour—every minute—there are new accounts that could be threatening me or somebody else to be raped and looking at child sexual images, but none of you are actually doing anything about it.

Nick Pickles

But the question you have here is, first, we are a global platform. There are people around the world right now who, if you say to them, “Can you go through that process and use your real name?” that would put them in physical harm. That might be because their religion is not tolerated in the country they are in. It might be because they are in Syria. It might be because the people who have been running the “Raqqa Is Being Slaughtered Silently” account have been murdered for running that account. I am not going to ask those people to put their real name to that account and those tweets. So yes, anonymity and not using real names is a challenge, but I also think that that is why we have rules. It doesn’t matter if you use your real name or not. There are lots of people who have accounts who don’t break our rules.

Naz Shah

I’m sorry, but that absolutely does not wash with me at all, and the reason for that is that you can have somebody that your organisation verifies but whose anonymity is still protected on the platform. That just doesn’t work for me at all.

Nick Pickles

I’m sorry, but there are lots of people out there who do not trust American technology.

Naz Shah

I am going to ask yes or no questions to all of you now. On YouTube, can you tell me what Google does when you can’t verify these accounts and you are putting up new ones all the time?

Peter Barron

In order to leave comments on a YouTube video, you have to be signed in. If you have a signed-in account and you are abusive and trolling on YouTube, you could be blocked or shut down.

Naz Shah

So you don’t verify YouTube accounts, either? And then you wonder why you’ve got ISIS and everybody else on there. You make $20 billion profit and you can’t even get the verification process right. Are you guys for real?

Peter Barron

The way that it works on YouTube is that you can—

Naz Shah

Sorry, but forget the content. Forget your standards and everything else; you can’t even get the basics right of growing your platforms with some corporate social responsibility and some ethics underlying all of that. That is what you are contributing to. Let’s be really clear. There are three outfits sitting before us—a Select Committee—giving evidence, yet none of you can give me the assurance that you will not have fake accounts going up while we are going through this evidence session, which could be promoting hatred, child pornography or extremism. None of you can give us that guarantee and yet you’re absolutely multinational, multi-billion pound companies. Then you have a look of shock and wonder why people are appalled at you.

Peter Barron

As Simon says, we would invite you to come and see the work of our trust and safety teams. I think you would genuinely be surprised, because you’ve just suggested that we don’t do anything in this area; the reality is very, very far from that. In 2015, our trust and safety teams removed 92 million videos from YouTube—

Naz Shah

But the horse has bolted, hasn’t it, Mr Barron? The reality is that you’ve still got accounts popping up every hour—every minute, more or less—on your platforms, and your concern here in all the evidence that you’ve placed before us is about growing the internet and growing the social platform, and it’s not about managing what you’ve already grown and actually being ethical in your future growth. None of you are telling me that, unless you can tell me something different.

Peter Barron

We work extremely hard on these issues. Everyone would accept that the opportunities, creativity and economic value that have been created by the internet are very significant. Now that is not to dispute that of course the platforms are used for negative purposes as well, but we are extremely dedicated to battling—it is a constant battle, and you were right to say that.

Naz Shah

What process do I have go through to get a Google account or a YouTube account?

Peter Barron

It is a constant, ongoing battle, as the web evolves, but in our trust and safety teams we have thousands of people and we are investing hundreds of millions of pounds in fighting acknowledged difficult issues that exist across the network.

Naz Shah

Are any of those hundreds of millions of pounds being invested to make sure people don’t even get on your platform in the first instance?

Peter Barron

Absolutely. We are extremely—

Naz Shah

Are they going through verification processes?

Peter Barron

On abusive accounts, we are—

Naz Shah

Sorry—not once they’ve got to the abuse. Before that. You can’t even control proscribed groups; you allow them platforms. What are you guys doing before they even get on your platform? What verification processes do you have to make sure that this is not an ISIS person and that this person is a bona fide taxpayer, or whatever, in this country—or whether they are even under age, or over age. You’ve got no way of knowing. My 12-year-old could have an account and pretend to be an adult, and you’d be all right with that, because you’ve got another user. That’s what’s coming across—

Nick Pickles

If we identified that, we’d remove their account, because our rules say they’re under 13.

Naz Shah

But they can get an email account without going through verification.

Nick Pickles

If you said, “My child is breaking the rules”, we’d take the account down.

Naz Shah

But when the horse has bolted—when the damage is done—you’re all happy to take the account down. You’re getting that wrong in the first instance. But what about the prerequisite for all of that, which is the verification of accounts? You’ve got children using accounts as adults, whether it’s Instagram, Facebook, Google or YouTube, and not one of your organisations has talked today about what you are doing prior, in terms of vetting, apart from putting people like us, for example, through loops and hoops to get verification ticks. Why hasn’t everybody got a verification tick? Why don’t we verify everybody?

Nick Pickles

Verification is open to apply for, and anyone can apply for it, and we ask them, as you say, to demonstrate their identity—

Naz Shah

But do you encourage it for every user?

Nick Pickles

I think that, given the social status it confers, people see it as something to aspire to.

Naz Shah

How many new Twitter accounts do you think have been opened today?

Nick Pickles

I don’t have those numbers to hand.

Naz Shah

Do you know how many have you taken down in the last 24 hours?

Nick Pickles

I don’t have those numbers to hand.

Naz Shah

In the last week, or in the last month?

Nick Pickles

I don’t have those numbers to hand.

Naz Shah

How many pages has Facebook taken down that have not been verified?

Simon Milner

I don’t have those numbers, but I’m very happy to answer your broader question, if you’d like me to.

Naz Shah

Yes, please.

Simon Milner

Much as we would all love to live in a country where all 65 million people were law-abiding, and I am sure that, as politicians who set the laws of the land, you would love to have a situation where everybody behaved themselves—we would like the same—they don’t all behave. That is why you—Parliament and Government—decide how much money should be spent on policing. We make those same decisions. This Committee deals with the most important issues for the country, in terms of keeping people safe. That’s also the most important issue for us. How do we keep people safe on our platform? We invest hundreds of millions of pounds, we employ thousands of people to work on this, and we are always trying to get better. I strongly disagree with your suggestion that we don’t care about this; we absolutely do. And we want to improve, and we want to be accountable for what we do—

Naz Shah

What do you do for verification, Mr Milner? Let’s go back to my original question.

Simon Milner

Sure. We do have a verification process. It is nothing to do with the safety of people on Facebook; it is about ensuring that you know whether you are looking at the actual page that Naz Shah MP is responsible for, or fans of her page or people who want to comment on what she is doing. It is about saying, “This is the authentic voice of Naz Shah.”

Naz Shah

How do you know it is authentic? You might have an email address that is not authentic.

Simon Milner

No, it is not that. We have a process. In the case of MPs, we have a specialist team that looks at—

Naz Shah

I’m talking about the average person on the street.

Simon Milner

The average person on the street cannot come to Facebook to be verified. That is not how Facebook works. Verification on Facebook is predominantly social, because you have to be your authentic self on Facebook—

Naz Shah

So, ISIS is its authentic self and you give it a page. Is that what happens?

Simon Milner

Would you allow me to finish? You have to be your authentic self on Facebook. You connect with people who are your friends, and most people—the vast majority—only connect with people they know in the real world. The way that they are authenticated is by those friendships.

Naz Shah

Mr Milner, I have just done a Facebook search and a Twitter search while we are taking evidence, and there are plenty of pages that are about banning Islam. There are plenty of right-wing pages on your Facebook platform. Whether they have 1,000, 2,000 or 3,000 likes, and whether it is right-wing or Islamic extremism, they peddle hatred. HT has platforms on Facebook and on Google and Twitter. Extremists have those platforms. At what point do you think it is not acceptable to have proscribed organisations and people who peddle hate on your platforms? What is the measure of your social responsibility in all of this? It appears to me that you are happy to grow your networks and your platforms, but you are not prepared to monitor them and make them ethical. That is how I feel.

Simon Milner

I am sorry that you feel like that, because it is fundamentally the opposite of my experience and the experience of many other people who are on Facebook and the organisations that we work with. We are working with lots of different organisations dedicated to combating hate speech as part of an EU process, right now, to test our systems and to figure out whether they work. On the whole, they are finding that we do have relatively robust systems, but there are certainly areas where we need to improve our knowledge about a particular type of hate speech or focus on a particular community in a particular European country. Through those processes, we learn. We are doing a lot already. We can do and are doing more, both using technology and working with organisations such as TellMAMA, which gave evidence to this Committee, and the police and others around the world—particularly in Europe, where this is a hugely important issue—to get better at this. We know that we have an important responsibility as European citizens and as European companies to help address this problem.

Naz Shah

Thank you very much. You referred earlier to us as lawmakers and the fact that everybody does not abide by the law. Who reviews your ethics and the laws that you have for your social standards? Do you have committees like this that review everything and take evidence from your social platforms? What happens?

Simon Milner

Our statute book is much smaller than yours. We have a set of community standards that are public and always have been. We try to make those as understandable as possible, so that people can really understand the nuance of particular issues. They are absolutely there for people to give us their views as to whether or not we are drawing the line in the right place. We also respect local law. There are some countries—not the UK—where the law goes beyond our standards. For instance, in Turkey it is illegal to defame Atatürk, the founder of modern Turkey, so in Turkey we have arrangements to ensure that, if there is that kind of content, we block it so that people cannot see it in Turkey. It is not illegal anywhere else in the world, so it is viewable everywhere else.

Naz Shah

So who actually makes those laws? We are here as lawmakers. We have a voting Lobby, and we table amendments. Who does yours? It is like a second world, isn’t it?

Simon Milner

Absolutely not.

Naz Shah

While we have our 65 million that you referred to, you have 1 billion on your platform.

Simon Milner

It is actually the real world. It is 1.9 billion people just like you and I and the people in this room, all around the world. The standards are the same everywhere. They are there for everyone to see. They were of course drawn up by a team of experts within our company, but we are absolutely ready and happy to take feedback on them. Just like every other organisation, when you use our service there are terms and conditions. Those are typically drawn up by the organisation and people decide whether or not they want to partake. That is true of every single company I can imagine that provides a service.

Naz Shah

Let’s just take that, for argument’s sake. Earlier, the Chair read from a piece of paper the line about Jewish people—with particular reference to Jews. In the law of this country, that would be anti-Semitic. Does our law have precedence over the law that you create to govern your sites? Which one takes precedence: the law of the land or—

Simon Milner

Local law, if local law goes beyond our standards. I can’t think of an example in the UK where we have had to act on content that is locally illegal here and is not against our terms. If there is an example, I will happily write to the Committee.

Naz Shah

Coming back to the statement the Chair referred to, about inciting hatred towards Jews, would you consider that anti-Semitic?

Simon Milner

Yes. I don’t think it was actually content on Facebook—it may have been one of the other platforms, but I may be wrong on that.

Naz Shah

It may have been Google. Mr Barron, if you would consider that statement anti-Semitic, why does it not breach your standards?

Peter Barron

We comply with the law in all the countries we operate in—

Naz Shah

You’re clearly not, because earlier you said you didn’t.

Peter Barron

And actually on YouTube, our community guidelines go above and beyond the law. Anything that was covered by UK law would be removed, either from Google Search or from YouTube. There is a legal requests form, easily found online, and if a user or anyone thinks that something that appears on YouTube is against the law, they can fill in that form and send it through, and our legal teams and policy experts will look at that and take a decision based on it. But if there was a court order saying that something was illegal, defamatory or copyright infringing, of course that would also come down.

Naz Shah

Mr Barron, for the final time—a yes or no answer—do you consider the following statement anti-Semitic? “Jews admit organizing White Genocide”. Yes or no?

Peter Barron

Of course it is.

Naz Shah

Right, but when the Chair asked you earlier whether that breached your standards, you suggested it did not, so which law do you actually apply?

Peter Barron

We apply the law of the country—

Naz Shah

Well, in this instance you clearly haven’t. Can you please refer back to us, as a Committee, and provide us with evidence that you have taken that content down because it is anti-Semitic? Can you do that for us?

Peter Barron

We will certainly come back to the Committee and explain our thinking.

Naz Shah

Thank you. I have a few more questions. Is child sexual exploitation automatically referred to the police in this country? While you are taking all this content down, is it automatically reported to the police?

Peter Barron

Yes.

Naz Shah

Is that a yes from all three of you?

Simon Milner

Just to be clear, it is not automatically referred to the UK police. We are under an obligation to report it to the US police, who then tip off the UK police. This is just a straightforward legal process. CEOP could give you evidence on this. We have to notify the US authorities, and they automatically notify the UK authorities, which can then get any information they need to get.

Naz Shah

If it is terrorist material, does that go to the FBI or to our counter-terrorism unit?

Simon Milner

It depends on who makes the request.

Naz Shah

So if you have something that is ISIS propaganda or a call for anything on your platforms, do you go to our counter-terrorism unit or our Met, or do you go to the FBI in America? Where do you report it?