A cybersecurity startup founded by former Department of Government Efficiency (DOGE) staffers reportedly penetrated sensitive internal systems belonging to artificial intelligence giant Anthropic in less than 12 hours, exposing vulnerabilities that raise serious questions about corporate cybersecurity and America’s technological defenses. The startup, Cathedral, used AI-powered agents to exploit weaknesses in third-party business applications, reportedly gaining access to internal meeting transcripts, employee compensation information, hiring records and communications involving advanced AI models. Anthropic acknowledged unauthorized access to a limited number of business systems but disputed the broader implications of Cathedral’s findings, maintaining that its core AI technology, source code and customer conversations remained secure. The incident has attracted attention within the Trump administration, where officials have expressed concern that foreign adversaries, particularly China, could exploit similar weaknesses to compromise strategically important American technology companies.
Key Takeaways
• AI-Powered Cybersecurity Threat: Cathedral researchers reportedly identified pathways into Anthropic’s sensitive internal systems in fewer than 12 hours using autonomous AI agents and vulnerabilities in third-party software.
• National Security Implications: Trump administration officials expressed concern that foreign intelligence services could exploit comparable weaknesses to obtain sensitive information from American AI developers.
• Corporate Accountability: Anthropic acknowledged limited access to internal business systems but denied that its core AI technology was compromised, exposing disagreements over cybersecurity responsibilities and the adequacy of third-party vendor oversight.
In-Depth
A cybersecurity investigation involving former DOGE employees has exposed troubling vulnerabilities in the business infrastructure supporting one of America’s leading artificial intelligence companies, raising questions about whether Silicon Valley’s cybersecurity protections can keep pace with increasingly sophisticated AI-powered attacks.
Cathedral, a startup established by former DOGE personnel, reportedly used autonomous AI agents to identify access routes into Anthropic’s internal systems in fewer than 12 hours. Researchers exploited weaknesses in third-party administrative software rather than directly compromising the company’s flagship Claude AI platform.
According to Cathedral’s findings, the exposed systems included employee compensation records, hiring information, internal meeting transcripts and corporate communications concerning advanced AI development. Anthropic acknowledged that researchers accessed portions of its sales, human resources and information technology systems but disputed the extent of the reported intrusion.
The company maintained that customer conversations, proprietary source code and AI model weights remained protected. Anthropic also stated that its monitoring systems detected suspicious activity and that affected software was disabled until vendors addressed the vulnerabilities.
The findings nevertheless attracted attention within the Trump administration, where officials reportedly expressed concern that foreign adversaries could exploit similar security weaknesses. China remains a particular concern because advanced American AI research represents an increasingly important strategic and economic asset.
Cathedral’s founders include former DOGE personnel Gavin Kliger, Luke Farritor, Marko Elez and Jack Stein. The company reportedly has secured federal cybersecurity contracts and substantial private investment.
The episode underscores a fundamental responsibility facing America’s technology industry: protecting strategically important intellectual property requires more than securing proprietary systems. It also demands rigorous oversight of outside vendors, transparent security practices and effective defenses against emerging AI-enabled threats.
Sources
• https://gizmodo.com/former-doge-staffers-new-startup-claims-it-hacked-anthropic-2000830560