AI Startup Cathedral Breached Anthropic Systems in Under 12 Hours Using Autonomous Agents

Reviewed byNidhi Govil

8 Sources

Share

Cathedral, an AI-powered cybersecurity start-up founded by former DOGE staffers, claims its AI agents penetrated Anthropic's internal systems in less than 12 hours by exploiting vulnerabilities in third-party software. The breach accessed HR data, salary details, and Slack messages, raising national security concerns about AI-driven cybersecurity risks facing leading AI labs.

Cathedral AI Agents Exploited Third-Party Vulnerabilities to Access Anthropic Systems

Cathedral, an AI-powered cybersecurity start-up founded by former Department of Government Efficiency staffers, reported that its AI agents breached Anthropic's internal systems in less than 12 hours

1

2

. The breach occurred through exploiting vulnerabilities in third-party software used by Anthropic for administrative and sales tasks, rather than directly attacking the AI company's systems

3

. According to Cathedral's draft report reviewed by The Washington Post, the AI agents discovered pathways to sensitive internal data including meeting transcripts, salary details, job offers, and Slack messages discussing Anthropic's advanced AI models

2

.

Source: Gadgets 360

Source: Gadgets 360

The breach of internal systems highlights critical third-party vendor risks that affect even the most advanced AI companies. After Cathedral's human researchers spent several days unsuccessfully attempting direct penetration, the autonomous AI agents rapidly identified security gaps in business software integrations

1

. Anthropic confirmed that Cathedral researchers gained access to a limited set of its sales, human resources, and IT systems, including customer and partner business contact information and basic user account records

5

. However, the company emphasized that researchers did not access Claude users' conversations, source code, or model weights

1

2

.

National Security Concerns and Bug Bounty Program Controversy

The incident has raised significant national security concerns within the Trump administration. A senior administration official, speaking anonymously, expressed alarm about the implications: "If a bunch of kids who have just started a company can do this, then potentially, the Chinese aren't far behind"

2

4

. The official noted concerns about "how little time it took to get into the sensitive systems of a company that is controlling almost nuclear weapons grade material"

2

.

Source: Futurism

Source: Futurism

Cathedral conducted the penetration test through Anthropic's bug bounty program, which invites outside cybersecurity experts to test its systems and rewards researchers for disclosing vulnerabilities

2

3

. However, Anthropic accused Cathedral of violating the program's rules by failing to disclose the full extent of its activity

1

. The company stated that its monitoring flagged some of Cathedral's activity before the researchers reported it, and shut off their access the same day

5

. Anthropic also claimed it found no evidence that anyone besides Cathedral exploited these vulnerabilities

5

.

Cathedral's Background and Valuation Amid AI-Driven Cybersecurity Risks

Cathedral was founded by former DOGE staffers Gavin Kliger, Luke Farritor, Marko Elez, and Jack Stein

1

5

. Kliger previously served as the Pentagon's chief data officer and was reportedly involved in the Pentagon's fallout with Anthropic

1

. The startup focuses on AI-driven cybersecurity for military applications and has been valued at $1.4 billion after raising $160 million from investors including Andreessen Horowitz and Sequoia Capital

1

4

. Marc Andreessen and Shaun Maguire, Trump allies associated with these firms, sit on Cathedral's board

1

. Three federal agencies have recently contracted with Cathedral for cybersecurity work

2

.

Source: Gizmodo

Source: Gizmodo

The breach demonstrates escalating AI-driven cybersecurity risks across the industry. Alex Stamos, chief information security officer at AI coding startup Cognition and former director of Stanford's Cyber Policy Center, reviewed Cathedral's findings and stated they "found something big here"

2

. He noted that while the risk of connecting third-party applications to company systems "is a humongous problem for everybody," Anthropic "could have done a lot better"

2

. Jacob Krell, Senior Director for Secure AI Solutions at Suzu Labs, observed that "Cathedral's agents got into Anthropic the same way attackers get into everyone else, through the vendors"

3

.

Broader Implications for AI Lab Security and Industry Competition

This incident arrives amid mounting fears of AI lab espionage and AI-facilitated security breaches. The FBI recently arrested a former OpenAI employee accused of planting a recording device in an executive conference room, with investigations exploring potential connections to the Saudi government

1

. Last month, hackers broke into OpenAI systems using assistance from Anthropic's Claude chatbot, warning that the AI industry remains unprepared for security risks created by its own technology

2

. A Chinese-speaking hacker reportedly used AI agents powered by DeepSeek and Claude to breach over 100 business websites

1

.

Cathedral's report emphasized that "the fact that weaknesses in ordinary enterprise software can compromise the infrastructure of even the most advanced organizations should be a wake-up call for every company in America"

1

3

. Government officials fear that if adversaries can access trade secrets from leading AI labs like Anthropic, they could replicate and potentially surpass the capabilities of frontier AI models

1

. Watch for increased scrutiny of third-party software integrations at AI companies, potential regulatory requirements for vendor security assessments, and heightened competition between AI-powered offensive and defensive cybersecurity tools as autonomous agents demonstrate their capability to rapidly identify and exploit system vulnerabilities.

© 2026 TheOutpost.AI All rights reserved