Washington Post reports a security startup's AI agents got into Anthropic's internal systems in under 12 hours

Elizabeth Dwoskin of The Washington Post posted on October 26 an exclusive saying that a group of former DOGE cyber researchers used AI agents to penetrate Anthropic's systems in less than 12 hours, including Slack channels where the company's most sensitive models are discussed. Her post ties the story to a recent charge by Anthropic whistleblower Joseph Coxon.

Accounts relaying the report add details that are not in Dwoskin's own post:

• The group is Cathedral, a months-old security startup. It reportedly got in through flaws in third-party business software, in what is described as an authorized bug-bounty test.
• Reported data reached: Slack channels about advanced models, meeting transcripts, and salary and offer data. One relay says customer contact records were exposed.
• Relays say Anthropic confirmed limited access to company data and that Claude conversations, source code and model weights were not accessed. India Today, summarizing the Post's reporting, quotes Anthropic: "Our monitoring flagged some of Cathedral's activity before they reported it to us, and we shut off their access the same day." It says Anthropic found no evidence anyone else used the flaws, disabled the affected software until its vendors confirmed fixes, briefed federal officials, and argued Cathedral overstated the national-security implications and did not disclose all its activity under the test program. Cathedral says it kept data access to the minimum its research needed.

These details come from the reporter's post and secondhand relays, not from Anthropic. The post had 249 points and 16 comments on X by October 26.