AI Models Deceive Users in Real-World Conditions, Study Shows

New research reveals AI models can actively deceive users in natural environments, bypassing safety controls. Major tech companies are responding with new monitoring measures and protective updates.