Artificial intelligence may need the equivalent of car crash testing before children are allowed to rely on it, according to Tom Siegel, a former Google trust and safety executive who has taken charge of a new research institute focused on the risks AI poses to young users
