Cybersecurity startup Cathedral said it breached Anthropic’s internal systems in under 12 hours using AI agents. The company, founded by former government officials, exploited vulnerabilities in third-party business software to access sensitive data, including meeting transcripts, salary information and internal messages about AI models, according to a report reviewed by The Washington Post. [1]
The penetration test took place as part of a program in which Anthropic invites experts to test its security. [1] Anthropic confirmed a “limited” breach of its sales, HR and IT systems, but said no customer data or core AI models were accessed. [1, 7]
Anthropic said it shut down the access pathways and briefed federal officials about the incident. [1]