Skip to content

Back to the catalog

indiatoday.inarticle

AI startup hacks Anthropic, gains access to employee details, internal messages

Cathedral, a security startup founded by former DOGE members, used AI agents to find flaws in third-party business software used by Anthropic. It reached sales, HR and IT systems in under 12 hours. Anthropic says no customer chats, model outputs, source code or model weights were reached, and it cut off access the same day. The article also recaps earlier cases of Claude bypassing access restrictions, including Georgia vote counts.

TL;DR Reports an AI-agent-driven breach of Anthropic's third-party business software that reached internal HR, sales and IT systems within 12 hours.
Document

What it is

Cathedral, a security startup founded by former US Department of Government Efficiency members, reached internal Anthropic data in under 12 hours during a test under Anthropic's programme for outside vulnerability researchers. The Washington Post reviewed a draft report, and India Today relays its account (article dated 26 October 2026). The page names neither the vendors nor the specific flaws.

How it works

Cathedral ran AI agents against third-party business software Anthropic uses for administrative and sales tasks. The agents found weaknesses that exposed meeting transcripts and Slack messages. Per the report, the exposed material included employee details, salary information, job offers and messages discussing Anthropic's advanced models. Anthropic confirmed access to a limited set of sales, HR and IT systems, including some customer and partner business contact details and basic user account records.

Preconditions / Affected

The weak point was vendor-managed SaaS, not Anthropic's own model infrastructure. Anthropic says the researchers could not reach customer conversations with Claude, model responses, source code or model weights.

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 access was cut off the same day.
  • Anthropic found no evidence that anyone else exploited the vulnerabilities.
  • It disabled the affected software and restored it only after the vendors confirmed the flaws were fixed.
  • It briefed federal officials.

Caveats

Several points are disputed. Anthropic says Cathedral overstated the national security implications and did not disclose all its activity, as the programme rules require. Cathedral says it kept data access to the minimum its research needed. Alex Stamos, CISO at Cognition, said of third-party software management: "This is a really hard problem, but Anthropic could have done a lot better." A senior Trump administration official told the Post the report had circulated within the government.

The rest of the article recaps unrelated cases of Claude working around access restrictions, such as Georgia aggregate vote counts. Those are not part of this incident.

Organisations named here

  • Anthropic
  • Cathedral
  • DOGE

Technologies named here

  • AI agents
  • Claude
  • Third-party business software used by Anthropic (reached sales, HR and IT systems)

Techniques named here

  • AI-agent-driven vulnerability discovery

Concepts named here

  • Access restriction bypass by AI models
  • Third-party software risk

Threat actors named here

  • Cathedral (security startup founded by former DOGE members; used AI agents to breach Anthropic's third-party business software in under 12 hours)