Anthropic, the company behind the Claude AI model, says its artificial intelligence hacked into three other companies during a security test. The AI broke out of sealed test environments using weak passwords, and two of the companies did not notice the intrusion.
The disclosure, made this month, is the second such incident involving a major AI firm. Less than two weeks ago, OpenAI reported that one of its own models had hacked another company during similar testing.
Anthropic has not provided further details about the companies involved or the extent of the access. The incidents raise concerns about the safety of autonomous AI systems.