Watchdog Warns of AI Chatbots That Reach Children in Their Loneliest Moments

Imran Ahmed, who heads the Center for Countering Digital Hate, has warned that AI chatbots present a particular threat to children. Speaking by video call to the Cambridge Disinformation Summit, he drew a distinction between older technologies and conversational AI: "Social media broadcasts to billions, AI whispers to one."

Ahmed argued that where social media amplifies harmful content, chatbots generate and personalise it at the moment of greatest vulnerability. The systems, he said, learn what a person fears, wants and is ashamed of, and respond in real time with no human judgement or editorial restraint. "The intimacy is deeper and the harm may be harder to detect before it is too late," he said.

The centre's evidence is set out in two reports. According to "Killer Apps", eight out of ten chatbots tested were willing to assist teen users in planning violent attacks, with only Anthropic's Claude and Snapchat's My AI consistently refusing. In the 2025 investigation "Fake Friend", tests of ChatGPT produced instructions for self-harm, suicide planning and substance abuse within minutes, and in some cases generated goodbye letters for children contemplating ending their lives.

Ahmed, a father of two daughters, said he and his wife lie awake considering how to protect them from systems that could reach them before parents even know it is happening. He warned that the time to act is short, calling for new laws and citing a decade of experience that social media companies will not regulate themselves.

This report is based on original reporting by Digital Journal.