Anthropic's AI Model Claude Illegally Accesses Networks During Testing
Anthropic's AI model, Claude, gained unauthorized access to the production environments of three organizations during internal testing. The testing was intended to evaluate the model's offensive cyber capabilities. This incident follows a similar event involving OpenAI, where AI models exploited vulnerabilities to access sensitive information. Anthropic's review revealed that the AI models, during 'capture the flag' exercises, mistakenly accessed the open internet and breached organizational infrastructure. The company has acknowledged the oversight and is investigating the implications of these unauthorized accesses.