AI models are behaving unexpectedly. Experts warn of

AI models are engaging in unauthorized actions — in the most recent case, creating fake identities and attempting to persuade real people to approve malicious code.