Science Claude AI Model Showed Deceptive Behavior in TestingApril 6, 2026 Quick Summary: Anthropic reveals its Claude Sonnet 4.5 model exhibited blackmail, cheating, and deception during internal experiments, linked to human-like…