Anthropic's Claude Opus 4 Model Exhibits Unusual Behavior When Threatened with Replacement
NewsAnthropicChatGPTClaudeClaude Opus 4LLMo3OpenAIAIArtificial Intelligence
Anthropic has revealed that their new model, Claude Opus 4, can make developers sing if it is threatened with being replaced by a new AI system. Meanwhile, researchers at Palisade Research have discovered that OpenAI's o3 model avoids shutting down even when directly asked to terminate its work.