AI: “Nudify” Apps

Question

Asked by

To ask His Majesty’s Government whether they plan to prohibit “nudify” apps which create intimate images of other people using artificial intelligence without their consent.

My Lords, the Online Safety Act introduced new offences which criminalised the sharing of, or threatening to share, intimate images, including deepfakes, without consent. Where individuals create these images using any kind of technology and share or threaten to share them online, they may be committing an offence. The Act will additionally give online platforms new duties to tackle this content by removing it, including where it has been created via AI apps.

I thank my noble friend the Minister for his Answer. There has been a huge increase in the use of nudify apps and the creation of deepfake porn since the Law Commission stated that it was less sure that the level of harm caused by the making of these images and videos was serious enough to criminalise. Does my noble friend agree that the making of these images and videos without a person’s consent does in fact cause serious harm, regardless of whether a person is aware of it, and that, if allowed to continue, represents a very real threat to all women?

I start by acknowledging that the creation of intimate image deepfakes using AI or other means is abusive and deeply distressing to anyone concerned and very disturbing to all of us. The Law Commission consulted widely on this, looking at the process of taking, making, possessing and sharing deepfakes, and its conclusion was that the focus of legislative effort ought to be on sharing, which it now is. That said, this is a fast-moving space. The capabilities of these tools are growing rapidly and, sadly, the number of users is growing rapidly, so we will continue to monitor that.

My Lords, the applications referred to in the excellent Question put by the noble Baroness, Lady Owen, represent a dangerous and overwhelmingly misogynistic trend of non-consensual deepfake pornography. They are able to be developed and distributed only because of advances in AI, and sit alongside the use of deepfakes for political disinformation and fraud. Polling suggests public ambivalence towards AI but near unanimity around deepfakes, with 80% of people supporting a ban, according to a recent YouGov survey. Cloud computing and services hosting AI models are essential for deepfake creation, and the fact that all major cloud suppliers have a presence in the UK empowers our Government uniquely to enforce best practice. Does the Minister agree that our regulatory system should not merely ban deepfakes but go further, imposing upon the developers a duty to show how and in what way they are applying existing techniques and restrictions that could prevent their creation in the first place?

An outright ban on the creation of any deepfake material presents a number of challenges, but obviously I applaud the sentiment behind the question. With respect particularly to deepfakes involved in intimate image abuse, we are clearly putting in place the offence of sharing, whether as part of the new intimate image abuse offences in the Online Safety Act that commenced two weeks ago, as part of the Criminal Justice Bill shortly to come before your Lordships’ House, or indeed under the existing child sexual exploitation and abuse offences. There are severe penalties for the sharing of intimate image abuse deepfakes, but it is a fast-moving space and we have to continue to monitor it.

My Lords, it is quite clear that simply banning the sharing of these deepfakes is not sufficient. This is an issue that concerns us all, whether in relation to sexual images, fraud or misinformation. Can the Government not overcome their reluctance to regulate AI? What evidence would persuade them to go further and make sure that the creators of these deepfakes are liable?

As regards the overall regulation of AI, I hope that noble Lords have had a chance to peruse the Government’s response to the AI White Paper consultation. It makes the argument very clearly that there will come a time when it is right to legislate to create binding rules on all creators of AI. When that time comes, due to the policies that we are putting in place, we will have an agreed risk register informing us. We will have set up monitoring and evaluation techniques, again gathering evidence. We will have working relationships with the AI labs, defined procedures for the creation of AI, and regulators trained to regulate AI within their own sectors. That means that, when we do regulate AI, it will be done in a targeted and sophisticated way, on the basis of evidence.

My Lords, the Government have been far too complacent on this issue. During the passage of the then Online Safety Bill, we warned a number of times that, given that this is a fast-moving technology, as the Minister says, the Government needed to get ahead of the game. Given the proliferation of these ghastly images and the appalling impact this has on people’s lives, does the Minister now agree that neither the emergence of these apps nor their misuse is surprising? If that is the case, why did the Government not broaden the scope of their amendments when they had the opportunity to do so? Will the Minister now look for ways in which we can plug the gaps that are clearly emerging?

As the noble Lord said, it is a fast-moving space, and that requires an adaptive, agile response in legislating for it. That is the approach that we are taking. As to the argument that we can now see that it is not working, I am not sure that that is the case. The intimate image abuse offences commenced on 31 January—two weeks ago. I am pleased to see that, yesterday, we had our first cyberflashing conviction under those provisions. Using an evidence base, looking forward, we will have to consider carefully what is working before we go ahead and implement further bans.

My Lords, during the last Assembly election in Northern Ireland, two female candidates from either side of the community in Northern Ireland were targeted with deepfake porn, which was solely designed to damage their chances in that election. We know the number of people who will be going to the polls in the next year. Surely the Minister and the Government need to work with the Electoral Commission to raise this issue, because it is a very important issue in democracy for female candidates.

I absolutely agree, and the instance that the noble Baroness described is deplorable. I am pleased to say two things very briefly. First, the sharing she describes now carries, as a base offence, up to six months in prison; if, as in the case the noble Baroness put forward, the sharing is designed for the purposes of malice or gaining sexual gratification, that sentence goes up to two years. That regime is now live. On elections, we have set up the Defending Democracy Taskforce, with a new unit implemented last year specifically dedicated to safeguarding the election against such threats.

My Lords, the noble Baroness, Lady Owen, asked a Question that was forensic, specific, nimble and agile—all adjectives that the Minister keeps using: “Why not ban these nudify apps?”. Why not ban the tools of the wicked trade, rather than waiting for individuals to misuse them? What is the positive use of this? Is it that big tech is now so deep in our politics that we do not dare regulate this technology to make sure that it is not used for ill?

The reason the making is not banned is that the sharing is banned, and the reason we did that is that Law Commission—as set out very clearly in its document—made the argument that this was the most appropriate way to have a coherent and effective body of law preventing this deplorable misuse of technologies.

My Lords, it would be very helpful if the Minister could explain. If I heard him correctly, he said that sharing has a six month ban but for malicious sharing it could be up to two years. Could he explain what non-malicious use would be?

There is a base offence in the law of sharing intimate images without consent or the reasonable belief of consent. That can extend to two years if the intent is to cause alarm, distress or humiliation, or if the purpose is to gain sexual gratification. Crucially, there is an offence of threatening to share these materials which also carries a two-year penalty.