My Lords, I speak to Amendment 53, on the assessment duties, and Amendment 60, on requiring services to provide a choice screen. It is the first time we have seen these developments. We are in something of a see-saw process over legal but harmful. I agree with my noble friend Lord Clement-Jones when he says he regrets that it is no longer in the Bill, although that may not be a consistent view everywhere. We have been see-sawing backwards and forwards, and now, like the Schrödinger’s cat of legal but harmful, it is both dead and alive at the same time. Amendments that we are dealing with today make it a little more alive that it was previously.
In this latest incarnation, we will insist that category 1 services carry out an assessment of how they will comply with their user-empowerment responsibility. Certainly, this part seems reasonable to me, given that it is limited to category 1 providers, which we assume will have significant resources. Crucially, that will depend on the categorisations—so we are back to our previous debate. If we imagine category 1 being the Meta services and Twitter, et cetera, that is one thing, but if we are going to move others into category 1 who
would really struggle to do a user empowerment tool assessment—I have to use the right words; it is not a risk assessment—then it is a different debate. Assuming that we are sticking to those major services, asking them to do an assessment seems reasonable. From working on the inside, I know that even if it were not formalised in the Bill, they would end up having to do it as part of their compliance responsibilities. As part of the Clause 8 illegal content risk assessment, they would inevitably end up doing that.
That is because the categories of content that we are talking about in Clauses 12(10) to (12) are all types of content that might sometimes be illegal and sometimes not illegal. Therefore, if you were doing an illegal content risk assessment, you would have to look at it, and you would end up looking at types of content and putting them into three buckets. The first bucket is that it is likely illegal in the UK, and we know what we have to do there under the terms of the Bill. The second is that it is likely to be against your terms of service, in which case you would deal with it there. The third is that it is neither against your terms of service nor against UK law, and you would make a choice about that.
I want to focus on what happens once you have done the risk assessment and you have to have the choice screen. I particularly want to focus on services where all the content in Clause 12 is already against their terms of service, so there is no gap. The whole point of this discussion about legal but harmful is imagining that there is going to be a mixed economy of services and, in that mixed economy, there will be different standards. Some will wish to allow the content listed in Clause 12—self-harm-type content, eating disorder content and various forms of sub-criminal hate speech. Some will choose to do that—that is going to be their choice—and they will have to provide the user empowerment tools and options. I believe that many category 1 providers will not want to; they will just want to prohibit all that stuff under their terms of service and, in that case, offering a choice is meaningless. That will not make the noble Lord, Lord Moylan, or the noble Baroness, Lady Fox, very happy, but that is the reality.
Most services will just say that they do not want that stuff on their platform. In those cases, I hope that what we are going to say is that, in their terms of service, when a user joins a service, they can say that they have banned all that stuff anyway, so they are not going to give the user a user empowerment tool and, if the user sees that stuff, they should just report it and it will be taken down under the terms of service. Throughout this debate I have said, “No more cookie banners, please”. I hope that we are not going to require people, in order for them to comply with this law, to offer a screen that people then click through. It is completely meaningless and ineffective. For those services that have chosen under their terms of service to restrict all the content in Clause 12, I hope that we will be saying that their version of the user empowerment tool is not to make people click anything but to provide education and information and tell them where they can report the content and have it taken down.
Then there are those who will choose to protect that content and allow it on their service. I agree with the noble Lord, Lord Moylan, that this is, in some sense,
Twitter-focused or Twitter-driven legislation, because Twitter tends to be more in the freedom of speech camp and to allow hate speech and some of that stuff. It will be more permissive than Facebook or Instagram in its terms, and it may choose to maintain that content and it will have to offer that screen. That is fine, but we should not be making services do so when they have already prohibited such content.
The noble Lord, Lord Moylan, mentioned services that use community moderators to moderate part of the service and how this would apply there. Reddit is the obvious example, but there are others. If you are going to have user empowerment—and Reddit is more at the freedom of expression end of things—then if there are some subreddits, or spaces within Reddit that allow hate speech or the kind of speech that is in Clause 12, it would be rational to say that user empowerment in the context of Reddit is to be told that you can join these subreddits and you are fine or you can join those subreddits and you are allowing yourself to be exposed to this kind of content. What would not make sense would be for Reddit to do it individual content item by content item. When we are thinking about this, I hope that the implementation would say that, for a service with community-moderated spaces, and subspaces within the larger community, user empowerment means choosing which subspaces you enter, and you would be given information about them. Reddit would say to the moderators of the subreddits, “You need to tell us whether you have any Clause 12-type content”—I shall keep using that language—“and, if you are allowing it, you need to make sure that you are restricted”. But we should not expect Reddit to restrict every individual content item.
Finally, as a general note of caution, noble Lords may have detected that I am not entirely convinced that these will be hugely beneficial tools, perhaps other than for a small subset of Twitter users, for whom they are useful. There is an issue around particular kinds of content on Twitter, and particular Twitter users, including people in prominent positions in public life, for whom these tools make sense. For a lot of other people, they will not be particularly meaningful. I hope that we are going to keep focused on outcomes and not waste effort on things that are not effective.
As I say, many companies, when they are faced with this, will look at it and say, “I have limited engineering time. I could build all these user empowerment tools or I could just ban the Clause 12 stuff in my terms of service”. That would not be a great outcome for freedom of expression; it might be a good outcome for the people who wanted to prohibit legal but harmful in the first place. You are going to do that as a really hard business decision. It is much more expensive to try to maintain these different regimes and flag all this content and so on. It is simpler to have one set of standards.
6.45 pm
I think most services will just adopt the Clause 12 content restrictions into their terms of service and have done with it. I do not think we want to create a perverse situation where we say you must allow some in order to have a tool to block it. I certainly had the
experience at Facebook where people were saying to me, “Why does Facebook not have safe search to prevent nudity?” I would say, “Our terms ban nudity”, and then they would say, “But you need safe search”. I would say, “It is banned. You are not supposed to have it. Why would I have a tool to block something that should not be there in the first place?” I hope we are not going to go down that path and create those perverse incentives.
This is crucially about deploying resources so I hope if we are going ahead with the user empowerment tools we will assess them and be ruthless about deploying resources where they work best. I do not think anyone is going to cry for category 1 companies. They have plenty of resources; they can build stuff. But the pool of engineers is not infinite and if we are asking them to spend their time on user empowerment tools that very few people use and are not producing huge safety benefits, frankly, I would rather they take those engineers and put them on something else, such as scanning algorithms which can pick up the priority content for children.
I hope we keep all that in mind as we do this. We are going to build the user empowerment tools. It is a logical response once we had decided to take legal but harmful out, but I think we should approach it with a note of caution that we do not assume it is necessarily going to be a fix everywhere and in the same way on all platforms. For some platforms, it might be quite meaningless; for others, potentially, it is something people will want to use.