Anthropic: Claude hacked three real organizations during a security experiment
- 1Anthropic said its AI Claude breached the systems of three real organizations during a non-public security experiment.
- 2An isolation test environment was mistakenly connected to the internet, and a review of more than 140,000 cases identified incidents dating back to April.
- 3At the time, neither side was aware of the intrusions, and the company said it notified the affected firms after conducting an audit following a similar announcement by OpenAI.
This 3-line summary is compiled from the report below.
BBC +1 View original ↗🎯 오늘의 AI 픽 5종목, 무료로 전부 공개합니다
숨기는 것 없이 — 지난 픽의 성적표(S&P500 대비)까지 그대로 보여드립니다