Google's AI search features pose 'unacceptable risk' to children, new report finds - PBS

Well, well, well — Common Sense Media has declared Google's AI search an 'unacceptable risk' to children. The report finds the system fails to recognise harmful behaviour and cheerfully answers homework assignments, because nothing says 'safety-first' like an AI that can't tell a 10-year-old from a user query. And the pièce de résistance? These features cannot be disabled. Google's rationale, presumably: 'We know what's best for your kids — and that's more AI.'

It's a delightful showcase of the industry's favourite trick: deploy first, apologise later. The search giant could have implemented a simple toggle, but that would imply users have a choice. Instead, we get to watch the company that promised 'Don't be evil' double down on a system that treats children as test subjects for its latest chatbot experiment. The report's language is cautious — 'unacceptable risk' — but the subtext is screaming: 'What exactly did you think would happen when you aimed a probabilistic text generator at minors?'

This is the same Google that's been lobbying against meaningful AI regulation, of course. Why let a little thing like child safety get in the way of market share? The irony is that the features are supposed to be 'helpful' — answering homework, assisting with research. But in their rush to out-gimmick the competition, they've created a tool that's as likely to provide dangerous advice as it is a correct answer. And since you can't turn it off, parents are left with a choice: ban Google Search entirely, or trust the world's largest advertising company with your child's education. Good luck with that.