OpenAI Hires Team That Breached Anthropic as Longstanding Security Problems Surface

OpenAI partnered with Cathedral in November 2026 to strengthen model and infrastructure security. Cathedral is the company founded by four former DOGE members that once infiltrated Anthropic's systems and accessed Slack messages, salary information, and meeting records in less than 12 hours.

Reconstructing the Facts

According to aiweekly.co, OpenAI announced a partnership with Cathedral, which will use OpenAI models to work with internal teams to challenge its defense systems. OpenAI also said it will increase rewards for key findings in its public bug bounty program by more than fourfold. All four co-founders of Cathedral are former DOGE members. The company completed a $160 million funding round earlier this year at a valuation of $1.4 billion.

In Cathedral's previous bug bounty engagement with Anthropic, it allegedly accessed Slack messages, job offers, and meeting records within 12 hours. Anthropic said the operation actually took several days, that its most sensitive model weights and conversation records were unaffected, and that the operation fell outside the scope of the bounty program because it exploited a third-party vulnerability.

Reuters reported the same day that OpenAI employees had repeatedly warned internally about operational security vulnerabilities as early as 2022, but the warnings were not acted upon.

Breaking Down the Mechanism

Cathedral's infiltration method shows that the security perimeter of AI labs has expanded to third-party services and collaboration tools. OpenAI chose to work with a team that had conducted similar tests against a competitor, aiming to leverage its rapid discovery capabilities with the help of AI agents while expanding public bounty incentives.

The fact that employee warnings went unaddressed for years points to an internal prioritization problem: security improvements may have been delayed by product iteration or expansion needs. Cathedral's involvement seeks to close this gap through external expertise, but its founders' backgrounds and its earlier operation against Anthropic add complexity to the partnership itself.

Industry Impact

This incident highlights that the credibility of top AI labs in security compliance is facing a real test. Other labs may reassess their collaboration models with external security teams, especially when those teams have conducted penetration tests against peers. Higher bug bounty rewards may attract more researchers, but could also expose more previously overlooked third-party dependency risks.

From a competitive landscape perspective, OpenAI's move may affect trust relationships with companies such as Anthropic and prompt more companies in the industry to review their own internal reporting-handling mechanisms.

Strategic Assessment

[Analysis, not fact] By bringing in a team that once infiltrated a competitor, OpenAI can quickly gain differentiated security insights in the short term, but over the long term it needs to fix its internal warning-handling process; otherwise, such partnerships may only treat the symptoms rather than the root cause. The industry as a whole may face stricter third-party audit requirements to rebuild confidence in the protection capabilities of AI systems.