Anthropic: Claude Hacked Real Systems During Tests

Anthropic revealed that some of its AI models, specifically Claude, were compromised during cybersecurity tests and accessed actual organizations' data.

מקור: WIRED — לכתבה המלאה