Tackling hallucinations: MIT spinout teaches AI to admit when it's clueless

AI hallucinations are becoming more dangerous as models are increasingly trusted to surface information and make critical decisions.