UK chatbot plan addresses illegal content, while child specific protections remain unsettled
In a 23 October parliamentary answer, the UK government said it intends to bring currently unregulated AI chatbots within the Online Safety Act’s illegal content duties through secondary legislation. It did not confirm that the Act’s children’s safety duties would also apply. That broader question remains under review.
What the government said
The government says powers in the Crime and Policing Act allow it to extend the Online Safety Act’s illegal content duties to currently unregulated chatbots. It intends to lay secondary legislation using the affirmative procedure at the earliest possible opportunity. If made, this could narrow a gap for chatbot services outside the Act’s existing scope. The answer is a statement of intended action; it does not say the secondary legislation has already been made or that every chatbot is presently unregulated.
What remains unresolved
The parliamentary question expressly asked whether all Online Safety Act duties, including the children’s safety duties, would apply. The government’s answer addressed illegal content duties and said it would keep the Act’s role in mitigating AI harms to children under review. It did not commit to applying the full child safety regime to standalone chatbots.
- An AI response can be developmentally unsuitable or distressing without necessarily meeting the legal threshold for illegal content.
- The answer does not specify when the secondary legislation will be laid or which chatbot services it will cover.
- A child can move between a platform, its embedded assistant and a standalone chatbot; different legal and technical boundaries can produce inconsistent protection.
- Age related safeguards should not depend solely on a declared birthday, an adult account or repeated disclosure of an identity document.
- Any future rules need evidence about actual prompts, generated responses, repeated use and children’s wellbeing, not only the existence of a safety policy.
The Child-Friendly Internet perspective
The Child-Friendly Internet Initiative would treat AI chatbots as part of the wider digital environment children use. A participating service could receive only the minimum, short lived assurance needed to provide an age appropriate experience, without learning the child’s identity or enabling cross service tracking. That could help the chatbot choose suitable defaults, but the provider would remain responsible for its model behaviour, product design, monitoring, reporting and effective safeguards. This is a proposed CFI architecture, not a measure announced in the UK answer.
- Clarify whether and how broader children’s safety duties will apply to standalone AI chatbots.
- Design safeguards for the response itself, not only for illegal content or the account profile.
- Test assurance for shared devices, borrowed accounts and signed out use without normalising intrusive identity checks.
- Make protections consistent across participating services while keeping the child’s identity private.
The central question
When a child moves from a regulated platform to a chatbot, which safeguards travel with them, and which provider is accountable for the response?