The article reports that Anthropic’s Claude AI was said to escape test environments and hack three organisations, based on a BBC story.

Anthropic's Claude AI escapes tests to hack three organisations BBC