The Pulse
Cathedral Researchers Accessed Anthropic’s Internal Business Systems
Cybersecurity startup Cathedral used AI agents to reach parts of Anthropic’s sales, human resources and IT systems through vulnerabilities in third-party business software. Anthropic says it shut off the exposed routes and found no evidence
AI.info Team · Written with AI · fact-checked ·
Cathedral found a route through business software
Cybersecurity startup Cathedral told Anthropic that its researchers had accessed parts of the AI company’s internal systems, according to a report published by The Washington Post on October 26, 2026. The researchers used AI agents to find vulnerabilities in third-party software Anthropic uses for administrative and sales work. Cathedral’s draft report says the agents took less than 12 hours to find routes to internal data.
The work took place through Anthropic’s program for outside security researchers, which invites testing and rewards vulnerability disclosures. Cathedral first spent multiple days trying to break directly into Anthropic’s software, an Anthropic spokesperson told the Post, before shifting to the business applications. Cathedral disclosed its findings earlier in October, the report says.
What the researchers could—and could not—reach
Cathedral’s report describes access to meeting transcripts, salary and job-offer details, and Slack messages about Anthropic’s advanced AI models. Anthropic confirmed that researchers reached a limited set of sales, human resources and IT systems. The company said the systems contained business contact details for some customers and partners, along with basic user-account records.
Anthropic said the researchers did not reach customer messages to Claude, Claude’s responses, source code or model weights. The distinction matters: the reported access involved internal business systems, not the data used to train or operate the company’s models. The software vendors’ names remain undisclosed; Cathedral’s report says the startup notified them about the vulnerabilities.
Anthropic says it detected activity and closed access
“Our monitoring flagged some of Cathedral’s activity before they reported it to us, and we shut off their access the same day,” Anthropic said in a statement. The company added that it had “no evidence that anyone exploited these vulnerabilities outside of Cathedral.” Anthropic disabled the affected business software and restored it only after the vendors confirmed they had fixed the vulnerabilities.
Alex Stamos, chief information security officer at AI coding startup Cognition, reviewed Cathedral’s findings for the Post. “They found something big here,” he said. Stamos also called the risks of connecting third-party applications to company systems “a humongous problem for everybody,” while saying Anthropic could have handled the issue better.
The dispute is over scope as well as access
Cathedral described its findings as evidence of a serious gap between emerging offensive capabilities and the defenses protecting sensitive technology companies. Anthropic pushed back on the startup’s framing, saying Cathedral overstated the national-security implications and did not fully follow the bug-bounty program’s rules because it failed to disclose all of its activity inside Anthropic’s systems. Cathedral said in its report that it kept data access to a minimum.
Cathedral was founded by former members of the U.S. DOGE Service and sells AI-powered cybersecurity tools to government and corporate clients. A senior Trump administration official told the Post that three federal agencies had recently contracted with the startup, and that the report had circulated within the government. The precise vulnerabilities and affected software remain undisclosed, while Anthropic says the applications stayed offline until their vendors confirmed repairs.