tl;dr: A study from French and Italian researchers found that giving people access to AI advice dropped their willingness to say "I don't know" from 44% to 3% and accuracy from 27% to 9%, while confidence jumped from 30% to 76%. Researchers deliberately used questions where the AI model (Step 3.5 Flash) typically failed, ruling out sensible delegation as an explanation. Monetary incentives to be accurate barely helped, suggesting the mere availability of AI erodes users' ability to recognize the limits of their own knowledge.