Ex-DOGE startup Cathedral breached Anthropic in under 12 hours, report says
What's the deal? Cathedral, a months-old cybersecurity startup founded by former US DOGE Service staffers, says its AI agents penetrated Anthropic's internal systems in under 12 hours, according to a draft report reviewed by The Washington Post. The researchers reached meeting transcripts, salary and job-offer details, and Slack messages discussing Anthropic's most advanced AI models.
How did they get in? Cathedral exploited vulnerabilities in third-party business software that Anthropic and many other companies use for sales and administrative tasks. It probed Anthropic under a company program that invites outside experts to test its systems.
What did Anthropic say? The company confirmed researchers reached "a limited set" of its sales, HR, and IT systems, but not customer conversations, model code, or model weights. It said Cathedral overstated the national-security angle of the test and didn't fully follow the rules of its bug-bounty program. Anthropic cut off the exposed pathways, disabled affected software pending fixes, and briefed federal officials.
Why now? The test lands amid heavy investment in AI-enabled cybersecurity. Reuters reported in July that Cathedral raised $160 million at a $1.4 billion valuation, with Andreessen Horowitz and Sequoia Capital backing plans to build military cyber capabilities. Cathedral co-founder Gavin KligerDealroom has a profile for this one. Try Dealroom → previously served as the Pentagon's chief data officer during the department's clash with Anthropic over restrictions on its models' use in federal systems.
What's the backdrop? Anthropic clashed with the Trump administration earlier this year after the Pentagon banned it from federal systems, following its refusal to allow its technology for autonomous weapons or mass domestic surveillance.
What could go wrong? One senior Trump administration official, speaking anonymously, warned that the ease of the breach pointed to wider risk. "If a bunch of kids who have just started a company can do this, then potentially, the Chinese aren't far behind," the official said.
The signal: The incident is less a compromise of Claude than a warning about the corporate systems surrounding high-profile AI developers. As companies adopt AI agents, attackers may exploit ordinary third-party software and access paths — not just the models themselves.
Investor reaction Josh Wolfe, co-founder of Lux Capital, described Cathedral on X as a "Lux family co" alongside a16z and Sequoia, and called the test "quite serious": critical to national security and to how "trillions of dollars of value at major labs and the frontier of research" must be protected.
Read more: The Washington Post, Digg, Josh Wolfe on X
