AI-Driven Cybersecurity Startup Cathedral Breached Anthropic in Under 12 Hours

Cathedral's Rapid Breach of Anthropic Raises AI Cybersecurity Alarms

Cathedral, an AI-driven cybersecurity startup valued at $1.4 billion, successfully breached Anthropic's internal systems in less than 12 hours using autonomous AI agents1

1

2

2

. The company, founded by former DOGE staffers Gavin Kliger, Luke Farritor, Marko Elez, and Jack Stein, demonstrated how exploiting vulnerabilities in third-party software can compromise even the most advanced AI labs. The incident has sent shockwaves through Washington, with senior Trump administration officials expressing deep concern about national security implications.

According to Cathedral's draft report reviewed by the Washington Post, the startup's AI agents discovered pathways to sensitive Anthropic data including meeting transcripts, salary details, job offers, and Slack messages discussing advanced AI models2

2

. The breach was conducted under Anthropic's cybersecurity testing program, which rewards researchers who identify software vulnerabilities, though Anthropic later accused Cathedral of violating program rules by failing to disclose the full extent of its activities1

1

.

What Cathedral Accessed and What Remained Secure

Anthropic confirmed that Cathedral researchers gained access to "a limited set" of its sales, human resources, and IT systems2

2

. The compromised data included business contact information for customers and partners, along with some basic user account records. However, Anthropic emphasized that the researchers did not access Claude users' conversations, source code, or model weights—the most sensitive intellectual property that defines the company's AI capabilities1

1

.

The AI company swiftly shut off the pathways exposed by the researchers and briefed federal officials on the incident2

2

. 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," while noting that managing third-party applications "is a humongous problem for everybody"2

2

.

Cathedral's Controversial Founders and Rapid Rise

Cathedral has raised $160 million from prominent investors including Andreessen Horowitz and Sequoia Capital, with Trump allies Marc Andreessen and Shaun Maguire sitting on the company's board1

1

. The startup is part of a wave of companies founded by former DOGE staffers seeking private sector opportunities and government contracts, with three federal agencies recently contracting Cathedral for AI cybersecurity work2

2

.

The company's leadership includes Gavin Kliger, formerly the Pentagon's chief data officer who was reportedly involved in the Pentagon's high-profile fallout with Anthropic1

1

. Co-founder Marko Elez had to resign from DOGE after being linked to racist tweets including "normalize Indian hate" and claiming he was "racist before it was cool," though Elon Musk promptly rehired him1

1

.

National Security Implications and Chinese Threat Concerns

A senior Trump administration official, speaking anonymously, revealed that Cathedral's report circulated within the government and raised significant national security concerns2

2

. "If a bunch of kids who have just started a company can do this, then potentially, the Chinese aren't far behind," the official stated, expressing alarm at "how little time it took to get into the sensitive systems of a company that is controlling almost nuclear weapons grade material"2

2

.

Government officials fear that if foreign adversaries exploit similar software vulnerabilities, they could access AI labs' trade secrets and potentially replicate or surpass the capabilities of advanced AI models1

1

. Cathedral's report emphasized this danger: "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

1

.

Growing Pattern of AI-Driven Hacks and AI Industry Security Challenges

The Cathedral breach follows a concerning pattern of AI-driven hacks and security incidents targeting leading AI labs. Recently, a Chinese-speaking hacker reportedly used AI agents powered by DeepSeek and Anthropic's Claude to break into more than 100 business websites over several weeks, targeting over 1,200 websites total1

1

. OpenAI was also recently targeted in an ethical hack that exposed software vulnerabilities affecting several employees' ChatGPT accounts1

1

.

Espionage concerns have intensified following the FBI's arrest of a former OpenAI engineer earlier this month for allegedly planting a recording device in an executive conference room2

2

. The Saudi national was allegedly in contact with a member of the Saudi royal family, with the FBI exploring whether he attempted to steal information to offer to the Saudi government1

1

.

Anthropic's Pentagon Tensions and Testing Program Dispute

The breach occurs against the backdrop of Anthropic's recent clash with the Trump administration. The Pentagon banned Anthropic from federal systems earlier this year after the company refused to allow its technology for autonomous weapons or mass domestic surveillance2

2

. This tension adds another layer of complexity to the security incident, particularly given Cathedral's connections to former government officials and current federal contracts.

While Anthropic noted that Cathedral researchers had previously spent several days unsuccessfully trying to breach its software directly before succeeding through third-party vulnerabilities1

1

, the company accused Cathedral of violating its cybersecurity testing program rules. A person close to Cathedral, speaking anonymously, said the researchers attempted to access Anthropic systems specifically to highlight the cybersecurity risks facing leading U.S. AI developers2

2

.