Recommendation 27

HM Government · verbatim extract Accepted

The government does not presently agree with the Committee’s recommendation for additional legislation in this regard. Doing so prior to full implementation of the OSA would complicate and undermine this process. The OSA lays the foundation for strong protections against illegal content and harmful material for children online, including content which is AI generated. AI generated content is already regulated under the OSA in the same way as ‘real’ content, i.e. where it is shared on an in-scope user-to-user or search service and constitutes either illegal content or content which is harmful to children. The largest user-to-user services will also need to enforce their terms of service, including in relation to AI generated content. Generative AI services that allow users to share content with one another or that search live websites to provide search results, are regulated under the online safety regime. In scope generative AI search or user-to-user services are required to assess the risk of harm to users from illegal content on their services and implement measures to manage and mitigate this risk. Where services are likely to be accessed by children, they will be required to take action to protect them from harmful content. Additionally, providers who publish pornographic content on their services must prevent children from accessing that content. This includes sites and apps that use, or allow users to use, Generative AI to generate pornographic material. Ofcom, as the independent regulator of the online safety regime, published an open letter on 8th November 2024 on how the OSA applies to Generative AI tools and providers. We continue to keep online safety under review. The government is committed to keeping young people safe online and we will act to address emerging AI harms. Nothing is off the table when it comes to keeping children safe. Alongside our work on online safety, the AI Security Institute (AISI) was set up to equip governments with a scientific understanding of AI’s risks so governments can act quickly to protect people’s safety and security. The Institute focuses on emerging AI risks with serious security implications, including the potential of AI to enable criminal activity. Additionally, AISI is investigating how highly capable AI systems can be used to manipulate, persuade, deceive, or subtly influence humans, and developing methods to measure these impacts. AISI recently published the first large scale study of its kind looking into the levers of political persuasion in conversational AI. AISI uses novel scientific methods to understand the risks of the most cutting-edge AI systems and is the first state-backed Institute in the world to have already conducted safety evaluations on frontier AI models. Following the launch of the UK’s AISI, we have seen similar organisations spring up around the globe – this is a clear testament to the UK’s approach. Through the AISI network and relationships with our international partners, the government is collaborating on how to tackle the shared challenges from AI. The Action Summit in Paris in February 2025 reinforced the Government’s commitment to international collaboration on the risks of frontier AI being used to generate mis/disinformation and threats online.