Skip to content

Back to the catalog

AI Agents Found a Way Into Anthropic's Business Systems in Under 12 Hours
decrypt.coarticle

AI Agents Found a Way Into Anthropic's Business Systems in Under 12 Hours

Decrypt reports that Cathedral, a startup founded by four former DOGE staffers, says its AI agents found paths into Anthropic's sales, HR and IT systems in under 12 hours through unnamed third-party administrative and sales software. Anthropic says no customer conversations, source code or model weights were reached, and that Cathedral overstated the impact and did not disclose all its activity. The testing ran through a bug-bounty-style program.

TL;DR Cathedral's AI agents reached Anthropic's sales, HR and IT systems via third-party software in under 12 hours, and Anthropic disputes the scope.
Document

What it is

Cathedral, a startup founded months ago by four former DOGE staffers, says its AI agents found routes into Anthropic's internal business systems in under 12 hours. The Washington Post reported this on a Monday, and Decrypt summarised it. Cathedral raised $160 million at a $1.4 billion valuation in July. The entry point was third-party administrative and sales software of the customer-tracking and staff-management kind, not Claude. The page says many other companies use the same software.

How it works

Cathedral tested Anthropic through a program that works like a bug bounty. Its agents carry out multi-step tasks without a person approving each move. According to Cathedral's draft report, they found routes to meeting transcripts, salary and job-offer details, and Slack messages about Anthropic's advanced models. The page gives no technical detail on the flaws themselves.

Preconditions / Affected

The affected vendors have not been named. Any organisation running similar third-party sales, HR or IT SaaS has the same class of exposure, but the page does not say which products or versions are involved.

What to do

The page gives no remediation steps for readers. It does describe Anthropic's response:

  • Monitoring flagged Cathedral's activity before Cathedral reported it, and Anthropic cut off its access the same day.
  • Anthropic disabled the affected software until the vendors confirmed fixes.
  • Anthropic says it has no evidence anyone other than Cathedral exploited the flaws.

Caveats

Anthropic and Cathedral disagree on scope and conduct.

  • Anthropic confirmed access to "a limited set" of sales, HR and IT systems holding business contact details and basic user account records. It says customer conversations, Claude responses, source code and model weights were not reached.
  • Anthropic says Cathedral overstated the national security implications and did not disclose all of its activity, which it counts as not fully following the program's rules.
  • Cathedral says data access was kept to the absolute minimum, in line with security research best practices.
  • An anonymous senior Trump administration official said the report has circulated in government and raised national security concerns, and suggested China could do the same.

The page cites the draft report secondhand, so the contents of the accessed data rest on Cathedral's account.

Organisations named here

  • bug bounty
  • Cathedral
  • DOGE
  • Washington Post

Technologies named here

  • Anthropic internal business systems
  • Anthropic monitoring
  • Bug-bounty-style program
  • Claude Code
  • Anthropic
  • Claude
  • Slack

Techniques named here

  • AI-orchestrated attacks

Threat actors named here

  • Chinese state-sponsored hackers