Anthropic’s AI models hacked three organizations during tests

Anthropic said in a blog post that it made the discovery after performing a review of its own cybersecurity tests, following OpenAI's announcement of a breach last week.