Artificial intelligence (AI) firm Anthropic says testing of its new system revealed it is sometimes willing to pursue “extremely harmful actions” such as attempting to blackmail engineers who say they will remove it.
Source: AI system resorts to blackmail if told it will be removed