Skip to content
UA.NEWS·Indexed by this site · Originally publishedSelectedAttention53

Cybersecurity startup Cathedral gained access to Anthropic data in under 12 hours

Cathedral不到12小时获取Anthropic数据访问权限

AI summary

US cybersecurity startup Cathedral, while permitted to test Anthropic's systems, gained access to some of Anthropic's internal data in under 12 hours through vulnerabilities in third-party commercial software. According to The Washington Post, researchers used AI agents to find ways into internal systems and obtained meeting notes, salary information, job offers, and Slack messages discussing advanced AI models. Anthropic confirmed that the access was limited to a limited set of sales, human resources, and information technology systems, said it had closed the access paths found and notified federal officials, and said Cathedral had previously spent several days trying to get into Anthropic's software. A senior Trump administration official said the report had circulated within the government, that such rapid access to sensitive systems raises concerns, and warned that similar weak points could be exploited by foreign intelligence agencies.

Why it was selected

Authorized researchers entered Anthropic's internal systems in under 12 hours via vulnerabilities in third-party commercial software, exposing the supply-chain attack surface of AI labs.

Read originalua.news