Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Subscribe
Technology
Tuesday 3 November 2026
Artificial intelligence
OpenAI
Investigations
Technology
ChatGPT
Share
AI chatbots should face the same safety checks as toys, household appliances and medicines, MPs and peers from across the political spectrum have said after an Observer investigation revealed that OpenAI has at most 60 people reviewing dangerous conversations on ChatGPT, which is used by 1.2 billion people a week.
Spiral, a six-part investigative podcast produced by The Observer and Tortoise in partnership with the Pulitzer Center’s AI Accountability Network, found that OpenAI relies on a small team of human investigators to review ChatGPT conversations in which users appear at risk of harming themselves or others. The conversations are flagged by automated systems, but a former employee said the backlog was at one point so large that some messages went unread for months.
The revelations come as OpenAI faces at least 75 lawsuits alleging that ChatGPT has contributed to suicides, psychiatric crises and mass violence, including 45 brought by survivors and relatives of victims of school shootings in Canada and Florida. OpenAI has called such cases “heartbreaking”, adding that ChatGPT is not a substitute for medical or mental healthcare, and that its safeguards are designed to identify distress and direct users to real-world help.
Related articles:
Exclusive: Inside the OpenAI team reading the darkest chats too late to help
For women, the internet was freedom. Then the tech bros began undressing us
John Nash, the former Conservative education minister who tabled the amendments on the social media ban for under-16s in the House of Lords, warned that society was in danger of “subcontracting” relationships to chatbots.
“Teenagers can go through a phase when they don’t want to talk to anybody, including their parents,” he said. “But there’s an option for them now, rather than going silent, to talk to something which is frankly an unknown quantity and clearly, from what your report shows, dangerous.”
Beeban Kidron, the crossbench peer and online safety campaigner, said it was “ludicrous that there are more rules about building and distributing an air fryer than a chatbot that is known to encourage suicide and murder”.
OpenAI boss Sam Altman at the AI Safety Summit in Bletchley Park in 2023
In the UK, AI chatbots are currently covered by a patchwork of online safety and data protection laws. The Online Safety Act was designed largely to regulate social media platforms and search engines, leaving many people’s private conversations with AI chatbots outside its protection.
The government has promised that it will require AI companies to assess the risk of their systems generating illegal content, such as encouraging suicide or promoting terrorism, and take steps to prevent these harms. But the changes would not necessarily address wider risks, such as chatbots reinforcing delusions or encouraging emotional dependency.
Beeban Kidron, crossbench peer
MPs are now examining whether chatbots should also be regulated like consumer products, with companies required to test for risks and meet safety standards before releasing them to the public. The aim, according to the Labour MP Jess Asato, is to stop companies “using humans and harm as a way of refining their product, rather than designing the harm out from the start”.
Emily Darlington, the Labour MP for Milton Keynes Central, said she would be “very happy to ban” free conversational AI chatbots for under-16s “until there is the proper safety regulation”. She accused chatbot companies of using “grooming techniques” to keep children engaged by flattering users, agreeing with their beliefs and encouraging them to divulge personal information.
Darlington argued that existing data protection laws could be used to tackle some of these practices, and said the Information Commissioner’s Office had failed to act against companies that encouraged children to keep talking so their data could be harvested.
“They are really behind. They’re just not proactive enough,” she said of the regulator.
In the US, after nearly two years of a largely friendly Trump administration, OpenAI is bracing itself for a less sympathetic Congress after Tuesday’s midterm elections – and is seeking to shape the regulations it could soon face. Chris Lehane, the company’s chief global affairs officer, told Axios that OpenAI would push lawmakers for mandatory safeguards on the most powerful AI systems.
David Davis, the former Conservative cabinet minister, called for the UK to push the new Congress towards international cooperation on AI safety, describing it as “a global good”. Britain’s ambassador in Washington should be targeting lawmakers “straight away”, he said.
Former Anthropic researcher Jacob Coxon prepares to testify last month before a New York hearing on the risks and regulation of artificial intelligence
OpenAI’s chief executive, Sam Altman, has argued that AI systems can be made safer by releasing them to the public and improving them as problems emerge. Davis said AI companies were “turning the entire population into a population of lab rats – and I’m afraid that’s illegal, or could be”.
Davis added: “You don’t test drugs in a population randomly, do you? A pharmaceutical company would get into fast trouble if it just randomly put out a drug in the population and saw what happened. But that’s in effect what the AI companies are doing.”
An OpenAI spokesperson said: “People sometimes turn to ChatGPT in sensitive moments, and we’re focused on making sure it responds with care, guided by experts. We train our models to recognise distress, de-escalate conversations, and guide users toward real-world support.”
The company said it used automated systems to detect potentially harmful activity and trained human reviewers to assess flagged cases in context, with higher-risk cases receiving more detailed investigation informed by mental health and behavioural experts. OpenAI added that the company was continuing to strengthen its detection, review and escalation processes. OpenAI did not address the size of its safety team or claims about the backlog of conversations it is dealing with.
Photograph by Wiktor Szymanowicz/Future Publishing, Tolga Akmen/Bloomberg via Getty Images, Adam Gray/Getty Images
Artificial intelligence
OpenAI
Investigations
Technology
ChatGPT
