Ex-Google safety chief warns AI poses deeper risk to children than social media

A former senior safety executive at Google has warned that artificial intelligence could inflict greater harm on children than social media did, urging the technology industry to slow its pace of development and accept outside oversight.

Tom Siegel, who founded and led the trust and safety team at Alphabet's Google, said the sector was repeating mistakes made during the social media era. He made the remarks as he announced that he had joined Common Sense Media, a children's online advocacy group founded in 2003, to serve as the first executive director of its Youth AI Safety Institute. The institute, launched in May, will assess how young people use AI. Anthropic and the OpenAI Foundation, the nonprofit that controls OpenAI, are among its backers.

Siegel worked at Google from 2004 to 2019 before leaving over what he described as the industry's plateauing ability to create safe online spaces. He later co-founded TrustLab, a startup that builds tools to detect safety threats, and stepped down as its chief executive earlier this month.

He said AI companies were releasing features for children without adequate guardrails, and that the risks ranged from suicide and psychosis to cognitive offloading — a decline in critical thinking caused by relying on AI for answers. "I think we're looking at very, very severe scenarios of harm being done," he said.

Siegel outlined steps he believes companies could take immediately. Anthropic and OpenAI could enforce stronger age verification to keep underage users off their chatbots, he said, while Google could add a simple option to disable AI Overviews and AI Mode in its search engine, similar to its existing toggle for filtering sexually explicit content. Such measures might affect usage or profits in the short term, he added, but firms that fail to act risk lawsuits and regulation that could hurt their businesses more.

A Google spokesperson said parents already have controls to turn off a child's access to the search engine entirely, adding that its AI search features are a useful way for young people to learn and explore. The company later said its AI tools undergo a rigorous testing and evaluation process focused on child safety. An OpenAI spokesperson said the company uses account and behavioural signals alongside self-reported age to identify underage users, and that it continuously tests and improves these systems. Anthropic did not immediately respond to a request for comment.

At the institute, Siegel said a top priority would be scaling up independent standards — comparable to car crash tests — to measure the risks of publicly available AI tools. Whether that succeeds will depend partly on AI companies' willingness to collaborate, which he said was uncertain but likely enough for him to take the role.