TL;DR
Parliamentarians from several parties say chatbots should meet safety standards before release, as medicines and household goods do, the Observer reported on 3 November. They were responding to its investigation, which says at most 60 people at OpenAI review dangerous ChatGPT conversations. OpenAI did not address that figure. Their argument points to consumer-product law as an alternative to online safety rules.
The Observer’s six-part podcast, Spiral, was made with Tortoise and the Pulitzer Center’s AI Accountability Network. Automated systems flag conversations in which users appear to be at risk, and a small human team then looks at them. A former employee told the programme the queue once grew so long that some messages sat unread for months. ChatGPT has 1.2 billion weekly users, the paper says.
OpenAI’s answer is that it trains models to recognise distress and guide people to real-world help, with higher-risk cases getting closer investigation informed by mental health experts. It says it is still strengthening detection and review. The Observer counts at least 75 lawsuits against the company, which has called such cases “heartbreaking”.
What the politicians said
Crossbencher Beeban Kidron called it “ludicrous that there are more rules about building and distributing an air fryer than a chatbot that is known to encourage suicide and murder”. Former Conservative minister John Nash warned of society “subcontracting” relationships to chatbots, and David Davis accused AI firms of turning the public into “a population of lab rats”.
Two Labour MPs went further on specifics. Jess Asato wants harm designed out rather than used to refine the product. Emily Darlington, who represents Milton Keynes Central, would be “very happy to ban” free conversational chatbots for under-16s until regulation is in place. She also says the Information Commissioner’s Office is “just not proactive enough” on data harvested from children.
The gap they are aiming at
The Observer describes current UK coverage as a patchwork, with the Online Safety Act built mainly for social media and search, so that many private chatbot conversations fall outside it. Ministers have promised to make AI firms assess and reduce the risk of illegal content, which we reported in February. The paper notes that this may not touch emotional dependency or reinforced delusions.
Product-style testing would be a different kind of rule, applying before release rather than after harm. MPs are examining it, but the report describes no bill or timetable. In our view, the near-term test is smaller: a separate letter on AI toys, also from 3 November, asks ministers to use consumer law on chatbots built into products. How they answer it will show whether the product-safety argument has any traction in government.