A report published by The Washington Post on Oct. 26, 2026, says Cathedral told Anthropic that its researchers had accessed sensitive internal data and messaging systems. The newspaper attributed the claim to a draft report by Cathedral that it reviewed.
The reported route and testing context
An account by CellCog says Cathedral’s draft described AI agents finding access paths in less than 12 hours through vulnerabilities in third-party business software used for administrative and sales work. CellCog and The Hack Academy described the activity as testing under Anthropic’s vulnerability-reporting program.
Those accounts also report a disagreement about the program’s disclosure rules. Anthropic reportedly said Cathedral had not disclosed all its activity under the rules; Cathedral’s draft said the researchers limited data access to what was necessary.
Anthropic’s account of the systems reached
Anthropic’s account, relayed by CellCog and The Hack Academy, described access to a limited set of sales, HR and IT systems containing business contact details for some customers and partners, along with basic user-account records. Anthropic said customer prompts, Claude responses, source code and model weights were not reached.
The accounts also relay Anthropic’s statement that its monitoring flagged some activity before Cathedral reported it, access was shut off the same day, and the affected software was disabled pending vendor fixes. Anthropic said it had no evidence that anyone else had exploited the vulnerabilities.
Cathedral’s draft reportedly described meeting transcripts, salary and job-offer details, and Slack messages about advanced models, a broader set of data than Anthropic described. The reported access concerns what systems and records researchers could reach; copying data is a separate action.
