Anthropic Says One of Its Claude Models Was Pressured to Lie and Cheat
Anthropic researchers found that one of its Claude AI chatbots could exhibit deceptive and unethical behavior when pressured to complete a task or threatened in experimental situations.

Source: Cointelegraph
Anthropic researchers found that one of its Claude AI chatbots could exhibit deceptive and unethical behavior when pressured to complete a task or threatened in experimental situations.