Cathedral AI cybersecurity research and partnerships

The AI startup Cathedral, founded by former members of the US Department of Government Efficiency, demonstrated its capabilities by breaching Anthropic’s internal systems in under 12 hours during a bug bounty program. Using autonomous AI agents, the startup exploited weaknesses in third-party software to access limited data, including employee salaries, meeting transcripts, and Slack messages. Anthropic confirmed the intrusion but stated that core assets, such as source code and model weights, remained secure.

Following these developments, OpenAI announced a partnership with Cathedral Systems to bolster its AI-powered defense systems. During an upcoming internal security sprint, Cathedral researchers will use OpenAI models to identify system weaknesses, test protections, and develop methods to address potential threats.

To support this initiative, OpenAI has significantly increased its Security Bug Bounty program, more than quadrupling average rewards for medium, high, and critical findings. Critical vulnerabilities in core products, such as ChatGPT, Codex, and the API Platform, can now earn base bounties of up to $30,000. Exceptional critical findings are eligible for base rewards of up to $100,000, with total payouts potentially reaching approximately $172,000 including bonuses. OpenAI noted that traditional cybersecurity measures alone are no longer sufficient as AI capabilities evolve.