# Ex-DOGE startup Cathedral says its AI agents got into Anthropic's internal systems in under 12 hours

Technology · 26 October 2026 · San Francisco, United States
URL: https://connectthedots.news/event/2026-10-26-ex-doge-startup-cathedral-says-its-ai-agents-got-into-anthropics

> Cathedral, a cybersecurity startup founded by former DOGE staffers, says AI agents reached Anthropic salary data, meeting transcripts and Slack messages through third-party software during an authorised bug-bounty test; Anthropic says access was limited and that Cathedral broke the programme's rules.

Cathedral, a months-old cybersecurity startup founded by former staff of the Department of Government Efficiency (DOGE), told Anthropic that its AI agents found routes into the company's internal business systems in under 12 hours, according to a draft Cathedral report reviewed by The Washington Post and published on Monday. Cathedral said the agents did not break Claude or Anthropic's own code but exploited weaknesses in unnamed third-party software used for administrative and sales work, reaching meeting transcripts, salary and job-offer details and internal Slack discussions about Anthropic's advanced models. The work was done under Anthropic's programme for outside security researchers, making it an authorised test rather than a criminal attack. [1][2][3]

Anthropic confirmed that the researchers reached what it called a limited set of sales, human-resources and IT systems holding business contact information for some customers and partners and basic user account records, and said customer prompts, Claude responses, source code and model weights were not accessed. It said its monitoring flagged some of Cathedral's activity before the startup reported it, that it cut off access the same day and disabled the affected applications until vendors shipped fixes, and that it found no evidence anyone outside Cathedral used the weaknesses. Anthropic said the researchers had earlier spent several days trying and failing to breach its software directly. [1][2][4]

The two sides dispute the findings' significance. Anthropic said Cathedral overstated the national-security implications and did not fully follow the programme's rules because it did not disclose all of its activity inside the company's systems; Cathedral said it kept data access to the minimum needed for responsible research and notified the software vendors, whose names it withheld. A senior Trump administration official told The Washington Post that three federal agencies have contracts with Cathedral and voiced concern that Chinese hackers could abuse similar flaws. Anthropic said it notified federal officials. [1][2][3][4]

Cathedral was co-founded by Gavin Kliger, Luke Farritor, Marko Elez and Jack Stein, all former DOGE staffers; Kliger previously served as the Pentagon's chief data officer and was reportedly involved in the Pentagon's dispute with Anthropic. Reuters reported in July that the company raised $160 million at a $1.4 billion valuation led by Andreessen Horowitz and Sequoia Capital, whose Marc Andreessen and Shaun Maguire sit on its board. [1][2]

The disclosure landed on a crowded day for Anthropic: the Justice Department filed a notice of appeal with the Ninth Circuit on Monday against Judge Rita Lin's August ruling that the administration had not justified barring federal agencies from using Anthropic's technology, Elvis Presley rights holders sued the company over the alleged use of 274 songs in training, and the Federal Trade Commission sought testimony from Anthropic executives in its probe of consumer harms from AI. Anthropic has also filed to list on Nasdaq under the ticker ANT. [4][5]

## Who was involved

- [Anthropic](https://connectthedots.news/org/anthropic): Confirmed limited access to sales, HR and IT systems; says core assets were untouched and Cathedral broke programme rules [1][2]
- [Cathedral](https://connectthedots.news/org/cathedral): Says its AI agents reached Anthropic's internal systems in under 12 hours via third-party software during an authorised test [1][2]
- [Gavin Kliger](https://connectthedots.news/person/gavin-kliger): Co-founded Cathedral; reportedly involved earlier in the Pentagon's dispute with Anthropic [1][2]
- [Andreessen Horowitz](https://connectthedots.news/org/andreessen-horowitz): Co-led Cathedral's $160 million round; Marc Andreessen sits on its board [1][2]
- [United States Department of Justice](https://connectthedots.news/org/united-states-department-of-justice): Filed a notice of appeal against the August ruling that blocked the ban on agencies using Anthropic [5]
- [Rita Lin](https://connectthedots.news/person/rita-lin): Issued the August ruling the government is appealing [5]

## How others covered it

Sides: Anthropic vs Cathedral. "Favours" is our reading of each article (see /sources#method).

- Gizmodo, 26 October 2026: "Former DOGE Staffers' New Startup Claims It Hacked Anthropic" [balanced] https://gizmodo.com/former-doge-staffers-new-startup-claims-it-hacked-anthropic-2000830560
- Bloomberg News, 26 October 2026: "Justice Department Appeals Ruling Allowing Anthropic AI Use by US Agencies" https://www.bloomberg.com/news/articles/2026-10-26/doj-appeals-court-ruling-lifting-ban-on-agency-anthropic-use
- Digital Trends, 26 October 2026: "Anthropic's sensitive internal data was exposed, and AI helped make it happen" [balanced] https://www.digitaltrends.com/cool-tech/anthropics-sensitive-internal-data-was-exposed-and-ai-helped-make-it-happen/
- Fox News, 26 October 2026: "Startup founded by ex-DOGE staffers used AI to access Anthropic systems in 12 hours: report" [balanced] https://www.foxnews.com/live-news/ai-news-claude-anthropic-10-26-26
- ZeroHedge, 26 October 2026: "Former DOGE Staffers Hacked Into Anthropic's Internal Systems In 12 Hours - And What They Found Is Alarming" [favours Cathedral] https://www.zerohedge.com/political/former-doge-staffers-hacked-anthropics-internal-systems-12-hours-and-what-they-found
- The Washington Post, 26 October 2026: "Researchers at cybersecurity start-up Cathedral got inside Anthropic" https://www.washingtonpost.com/technology/2026/10/26/researchers-cybersecurity-start-up-cathedral-got-inside-anthropic/
- Cybersecurity News, 26 October 2026: "Researchers Use AI Agents to Breach Anthropic's Internal Systems in Just 12 Hours" [balanced] https://cybersecuritynews.com/anthropic-internal-system-breach/

## Linked stories

- Related to: [Anthropic's draft IPO filing shows an $80 billion first-half loss as investors weigh a lower valuation](https://connectthedots.news/event/2026-10-21-anthropics-draft-ipo-filing-shows-an-80-billion-first-half-loss-as) (21 October 2026)

## Sources

1. Gizmodo, "Former DOGE Staffers' New Startup Claims It Hacked Anthropic", 26 October 2026: https://gizmodo.com/former-doge-staffers-new-startup-claims-it-hacked-anthropic-2000830560
2. Cyber Security News, "Researchers Use AI Agents to Breach Anthropic's Internal Systems in Just 12 Hours", 26 October 2026: https://cybersecuritynews.com/anthropic-internal-system-breach/
3. Digital Trends, "Anthropic's sensitive internal data was exposed, and AI helped make it happen", 26 October 2026: https://www.digitaltrends.com/cool-tech/anthropics-sensitive-internal-data-was-exposed-and-ai-helped-make-it-happen/
4. Fox News, "AI company Anthropic sued over alleged Elvis Presley copyright infringement", 26 October 2026: https://www.foxnews.com/live-news/ai-news-claude-anthropic-10-26-26
5. Bloomberg, "Justice Department Appeals Ruling Allowing Anthropic AI Use by US Agencies", 26 October 2026: https://www.bloomberg.com/news/articles/2026-10-26/doj-appeals-court-ruling-lifting-ban-on-agency-anthropic-use
