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.

Original source: https://www.cbsnews.com/news/ai-models-behaving-unexpectedly-security-experts/