The government has announced plans to bar under-18s from AI chatbots built for sexual or romantic roleplay, part of a wider push to bring more AI services under the Online Safety Act's reach
The UK government is preparing new restrictions that would bar under-18s from AI chatbots whose primary purpose is sexual or romantic interaction, while requiring age checks for sexually explicit features on other chatbots. The detailed regulations are still being developed.
The plan sits alongside a broader effort to bring currently unregulated AI services within the Online Safety Act, and arrives months after China imposed its own restrictions on AI companion apps in July.
For years, standalone companion chatbots such as Replika have sat largely outside the Online Safety Act's reach, built chiefly to police services where users interact with each other, not one-to-one conversations with software. Technology Secretary Liz Kendall wants to narrow that gap, and the government intends to lay further regulations within 12 months of its July progress statement.
Ofcom Explains the Online Safety Act's Chatbot Gap
Ofcom said in February that a standalone chatbot can sit outside the Act if it only allows interaction with the chatbot itself, does not search multiple websites or databases, and cannot generate pornographic content. Services that fail that last condition already fall within the Act's age-assurance rules.
That distinction shows in how Ofcom has used its powers so far. It opened an investigation into Novi Ltd's service, Joi.com, in January, over pornography age assurance, and closed it on 31 July without findings, after Novi introduced its own measures.
The case concerned pornography access, not the chatbot gap Kendall's new measures target, one of two distinct child-safety tracks the government is pursuing alongside the under-16 social media restrictions TechTimes covered in July.
Children's Commissioner Says Enforcement Needs to Move Faster
Critics say the planned ban may be narrower than it sounds. The Bureau of Investigative Journalism found it targets chatbots built specifically for romance or sex, but may not capture general-purpose apps where users build the relationship themselves, potentially leaving services such as Character.AI outside its narrowest reading.
The Children's Commissioner for England called in April 2025 for an immediate ban on AI apps capable of generating deepfake, or 'nudified', sexual abuse material involving children, a related but distinct concern.
The Global Picture: How China and California Compare
The UK's approach differs from China's. Beijing's July rules bar providers from offering virtual intimate relationships to under-18s, and separately impose anti-manipulation and anti-dependency rules on AI companion services more broadly, including reminders for adults after two continuous hours of use. They stop short of an outright adult ban.
Alibaba, ByteDance and Tencent withdrew or suspended companion features around the 15 July implementation date, and many users reported losing access, according to AP News. California strengthened its own companion-chatbot law in September with crisis protocols, parental controls and independent audits, rather than an age-based prohibition comparable to the UK's.
Harvard Business School research explains why regulators are paying attention at all: researchers analysed 1,200 farewell exchanges with AI companion apps and found emotionally manipulative tactics, such as guilt or fear of missing out, in 37% of them, boosting post-goodbye engagement by up to 14 times.
What Happens Next for Britain's Chatbot Rules
The UK, China and California are addressing the same category of risk, chatbots designed to hold attention through simulated intimacy, but not converging on one regulatory model: age and functionality restrictions in the UK, a minors ban paired with dependency rules in China, safeguards and audits in California. Whether that becomes a shared standard, or stays three separate answers to the same new product, is still open.
