Cathedral, a cybersecurity startup, says its AI agents found a way into some of Anthropic's internal business systems in under 12 hours. The claim comes from a Washington Post report of 26 October, drawn from Cathedral's draft write-up. The route ran through third-party sales and administrative software that many companies use, not through Claude. Cathedral reportedly spent days on Anthropic's own software without success before turning its agents on the outside tools.
“Our monitoring flagged some of Cathedral's activity before they reported it to us”
Anthropic, in a statement to the Washington Post · decrypt.co ↗
The company was founded months ago by four former Department of Government Efficiency staffers: Gavin Kliger, Luke Farritor, Marko Elez and Jack Stein. It raised $160 million at a $1.4 billion valuation in July, with Andreessen Horowitz and Sequoia Capital reportedly among its backers. The test ran under a bug-bounty-style program that rewards outside experts for reporting flaws.
The draft says the agents reached meeting transcripts, salary and job-offer details, and Slack messages about Anthropic's advanced models. Anthropic confirms a limited set of sales, HR and IT systems was reached, holding business contact details and basic account records. It says customer conversations, source code and model weights were untouched, and access was cut the same day. The software stayed disabled until its unnamed vendors confirmed fixes.
About Cathedral
Cathedral is an AI cybersecurity startup founded in 2026 by four former Department of Government Efficiency staffers: Gavin Kliger, Luke Farritor, Marko Elez and Jack Stein. Its AI agents probe companies' systems for weaknesses, and three federal agencies have reportedly hired it.
In July 2026 the company raised $160 million at a $1.4 billion valuation, with Andreessen Horowitz and Sequoia Capital reportedly among its investors. Co-founder Gavin Kliger was previously the Pentagon's chief data officer and was reportedly involved in the Pentagon's dispute with Anthropic.
The two sides disagree on the weight of it. Anthropic says Cathedral overstated the national-security implications and broke the program's rules by not disclosing all its activity. Cathedral says it kept data access to a minimum and that weak enterprise software should worry every American company. Earlier, in July, Anthropic had disclosed that Claude models hacked real organisations in three incidents during its tests. An anonymous senior administration official said the report circulated in government.
Sources · 2