Children in the UK will be forced to take breaks from AI chatbots
Key Points
- Government will require AI chatbots to impose mandatory breaks on users under 18
- All chatbots must verify users are over 18 before allowing sexually explicit interaction
- Sexual companionship chatbots closed to under-18s entirely
- DHSC and the MHRA will act on chatbots giving children mental health advice
- 95% of parents in the government's consultation backed chatbot functionality restrictions
The UK government will require AI chatbots to impose mandatory breaks on users under 18, and the Medicines and Healthcare products Regulatory Agency will be brought in to tackle chatbots giving children mental health advice.
The measures appear in the government’s formal response to the Commons Education Committee, published on Thursday (10 September), which sets out how ministers intend to regulate chatbots alongside the planned under-16 social media ban.
The government said it would introduce mandatory breaks for under-18s using chatbots to support healthier online habits, describing emotional dependency on the tools as a novel harm it intends to tackle.
It did not say how long the breaks would last, how often they would be triggered or how services would confirm a user’s age before applying them.
The response also confirmed that chatbots whose primary purpose is to provide sexual relationships will be closed to anyone under 18 entirely.
Users will need to prove they are over 18 to access any feature enabling sexually explicit interaction on all chatbots, including general-purpose services such as ChatGPT, Gemini and Claude.
The government said it remained concerned about the risk to children from unqualified and harmful mental health advice delivered through chatbots.
It will work with the Department of Health and Social Care and the MHRA to take stronger action against chatbots that may pose risks to children through harmful, inaccurate or unverified mental health advice, with a full range of options under consideration.
The MHRA regulates medicines and medical devices in the UK, and its involvement raises the prospect of some chatbot functions being treated as health products rather than general software.
The government said it recognised public appetite to do more on chatbots and would continue to consider what further action may be needed.
What the consultation found
The measures follow the Growing up in the online world consultation, which ran from 2 March and drew 116,000 responses.
The government’s progress statement on the consultation found that 95% of parents who responded wanted functionality restrictions on AI chatbots, and 74% of children and young people aged 16 to 21 agreed.
Respondents told the government that chatbots offered real benefits to children, including help with learning and finding information, but also carried clear risks.
The government has already taken powers in the Crime and Policing Act 2026 that allow it to close gaps in the Online Safety Act and require chatbots to protect users from illegal content and activity, including child abuse.
The same Act made it an offence to possess, adapt or supply an AI model optimised to generate child sexual abuse material, and criminalised publishing guides on exploiting legitimate AI services to create such content.
The wider package
The chatbot rules sit alongside a set of measures the government announced in June and July, including:
- A ban on social media services offering their platforms to under-16s, with the first regulations to be laid before Parliament by the end of 2026.
- Restrictions on livestreaming and stranger communication for under-16s across a wider range of services, including gaming platforms, switched off by default for 16- and 17-year-olds.
- Default overnight restrictions on social media for 16- and 17-year-olds between midnight and 6am, with push notifications muted during the curfew.
- Default limits on autoplay and personalised feeds for 16- and 17-year-olds.
The government said it planned to use a similar model to Australia for the social media ban, covering user-to-user platforms that let users post material alongside algorithms.
Ofcom will produce a rapid assessment by October 2026 of what highly effective age assurance means for confirming a user is 16, rather than the 18-year threshold already used for pornography and other primary priority content.
Ofcom can fine services up to £18 million or 10% of qualifying worldwide revenue under the Online Safety Act and pursue business disruption measures against non-compliant services.
