UK Moves to Hold AI Chatbot Makers Legally Liable for Illegal Content

Prime Minister Keir Starmer announced plans on Monday to amend the Online Safety Act, closing a legal loophole that has exempted generative AI chatbot providers from illegal content duties for more than two years. The amendment would subject all AI chatbot operators in the UK to the same statutory obligations as social media platforms, with penalties that include fines of up to 10 per cent of global revenue and the possibility of court-ordered service blocks within the United Kingdom. The policy shift follows intense public backlash after Elon Musk's Grok AI assistant generated non-consensual sexualised deepfakes of real people, an episode that prompted X to stop its Grok tool from creating such images in the UK last month. Starmer said the government had prevailed in its confrontation with X over Grok, but that the same standards must now apply across the entire AI chatbot sector. The chief executive of the NSPCC welcomed the direction of the changes, which represent one of the most direct government attempts to regulate synthetic media generation at the model level rather than merely policing distribution.

Current online safety legislation primarily covers content shared between users on platforms rather than outputs generated by AI models themselves. If AI chatbots are used specifically as search engines, to produce pornography, or operate in user-to-user contexts, they are already covered by the act. However, they can be used to create material that encourages people to self-harm or take their own lives, or even generate child sexual abuse material, without facing sanction. Ofcom, the UK's communications regulator, admitted it lacked powers to act against Grok because images and videos created by a chatbot without it searching the internet are not in the scope of the existing laws, unless they amount to pornography. The expanded duties would require AI developers to implement systems preventing their models from producing illegal material, including child sexual abuse content and non-consensual intimate imagery. Ofcom would gain enforcement authority over both closed and open-source models operating in the British market, though technical questions remain about how compliance will be verified for models deployed across international borders, particularly open-weight systems that can be downloaded and run locally.

The government said it would move fast to shut the loophole, with the change potentially taking effect within weeks, as more and more children use chatbots for everything from help with their homework to mental health support. Starmer stated that technology is moving really fast and the law has got to keep up, adding that the action taken on Grok sent a clear message that no platform gets a free pass. The announcement came alongside an accelerated consultation on potential social media bans for users under 16, restrictions on addictive design features such as infinite scrolling, tighter VPN controls, and mandatory data preservation requirements for tech companies when a child dies. Australia's recent legislation banning users under 16 from major platforms has provided a reference point, though UK ministers have not committed to a blanket ban. The consultation will also examine curbs on algorithmic recommendation systems and infinite scroll mechanisms that drive prolonged engagement among minors. Additional proposals include restrictions on VPN usage to circumvent age verification and requirements for platforms to preserve user data when a child's death triggers an investigation. Starmer indicated that any changes to children's use of social media, which may include other measures such as restricting infinite scrolling, could happen as soon as this summer.

The Conservative opposition dismissed the government's claim to be acting quickly as more smoke and mirrors given the consultation has not yet started. Laura Trott, the shadow education secretary, said claiming they are taking immediate action is simply not credible when their so-called urgent consultation does not even exist. She added that Labour have repeatedly said they do not have a view on whether under-16s should be prevented from accessing social media, which she described as not good enough, and that she is clear that under-16s should be stopped from accessing these platforms. Opposition politicians have welcomed the direction on AI regulation but questioned whether the consultation timeline is realistic given the technical complexity of enforcing duties on globally distributed AI systems. The government maintains that the Grok episode demonstrated the urgency of action and that the existing Online Safety Act framework provides the necessary enforcement architecture. Parliament will need to approve the amendments before they take effect.

The package marks a regulatory pivot from holding platforms accountable for user-generated content to imposing direct legal liability on the builders of generative AI systems. It also tests the limits of state intervention in product design choices that have historically been treated as commercial decisions. Civil liberties groups have raised concerns about privacy implications of device-level scanning and VPN restrictions, while technology companies argue that broad liability for model outputs could stifle innovation and drive development offshore.