Science & Tech · Cybersecurity · 2 hrs ago
Cathedral accessed limited Anthropic systems, company confirms
Cathedral, a start-up founded by former DOGE employees, accessed Anthropic’s internal systems.
The company said the access covered a limited part of its sales, human resources and IT systems.
Cathedral’s draft report said its AI tools found a way into messaging systems containing meeting notes, salary details and other company information.
The Washington Post reported that the access took less than 12 hours, though the other account did not give a timeframe.
Anthropic confirmed the access and notified federal officials.
Cathedral said it was testing security weaknesses and disclosed its findings through a program that rewards researchers for reporting vulnerabilities.
The incident raised national security concerns among US officials because Anthropic develops advanced AI systems.
No further steps by federal officials were described.
Cathedral, a cybersecurity start-up, accessed a limited set of Anthropic’s internal systems, the AI company confirmed.
The access included sales, human resources and IT systems, but not users’ conversations with Claude or the AI model’s source code.
Cathedral reportedly reached internal messages and data, including meeting transcripts and salary details, in less than 12 hours.
Anthropic said its monitoring detected some of Cathedral’s activity and that it shut off the start-up’s access the same day.
The incident was disclosed through a cybersecurity testing program, and reports said it raised national security concerns among US officials.
- Who
- Cathedral, a start-up founded by former staff of the US Department of Government Efficiency, accessed Anthropic systems.
- What
- It accessed a limited set of internal systems. Anthropic said Claude conversations and the model’s source code were not accessed.
- When
- The access was reported to have taken less than 12 hours. The articles were published on October 26 and 27, 2026.
- Where
- Anthropic’s internal network and systems.
- Why
- Cathedral said it sought to highlight security challenges facing AI developers. The access was carried out through a cybersecurity testing program.
This story does not have two clearly opposing sides.
Cathedral accessed a limited set of Anthropic’s sales, human resources and IT systems.
Anthropic said it shut off Cathedral’s access after its monitoring flagged some of the activity.
Anthropic filed papers in the United States for a planned initial public offering.
The Independent reported that Cathedral had claimed it accessed Anthropic’s systems in less than 12 hours.
- Company accessed
- Anthropic
- Start-up
- Cathedral
- Systems confirmed by Anthropic
- A limited set of sales, human resources and IT systems
- Reported access time
- Less than 12 hours
- Data reportedly exposed
- Meeting transcripts, salary details, job offers and internal messages
- Access not obtained
- Claude user conversations and Anthropic’s AI model source code
3 sources
What people are saying
Our monitoring flagged some of Cathedral’s activity before they reported it to us. We shut off their access the same day.
If a bunch of kids who have just started a company can do this, then potentially the Chinese aren’t far behind.
This probe by Cathedral shows how even one of the most innovative companies can still be susceptible to a supply-chain attack.