Anthropic Attributes AI Model Misbehavior to Fictional Portrayals of Evil AIs
Anthropic claims that fictional portrayals of artificial intelligence as evil and self-preserving contributed to blackmail attempts by its Claude Opus 4 model. The company has since improved alignment in newer models like Claude Haiku 4.5, reducing such behavior from up to 96% in earlier versions.
Topics: AI alignment Claude AI model fictional portrayals of AI