OpenAI expands security testing with Cathedral Systems

OpenAI is planning a new internal security sprint with Cathedral Systems, bringing the firm’s researchers together with OpenAI teams to use OpenAI models to test and strengthen the company’s systems. OpenAI also says it will more than quadruple rewards for critical findings in its Security Bug Bounty program.

The initiative is intended to bring outside security expertise into an ongoing effort to identify weaknesses, test fixes and harden OpenAI platforms. Cathedral Systems researchers will work alongside OpenAI teams while using the company’s models to challenge defenses with differentiated research approaches.

External research and model-assisted testing

Gavin Kliger, CEO and co-founder of Cathedral Systems, said the collaboration combines OpenAI’s models and security expertise with Cathedral Systems research to find weaknesses, test fixes and strengthen continuous defense. OpenAI said it wants researchers to challenge its assumptions and improve both its systems and its defensive methods.

The expanded bounty rewards are part of that approach. By increasing incentives for critical reports, OpenAI is seeking additional scrutiny of the systems and processes it uses to protect its platforms.

Building on the last security sprint

OpenAI said its previous sprint assigned 25% of product engineers to finding and fixing weaknesses. Using Codex, those engineers investigated reported findings, reproduced vulnerabilities in isolated environments and prepared patches for review.

The company frames the next sprint as a response to a changing threat environment in which traditional cyber defenses alone are no longer sufficient. Its objective is to build continuous defensive loops that find and fix weaknesses before attackers do.

That broader program includes making its most capable models available to Daybreak customers. The company has committed $1 billion in subsidized access over six months for frontline defenders protecting essential services and digital infrastructure. Its work to support open-source security also includes OpenAI support for open-source maintainers, which focuses attention on the maintainers responsible for widely used software.

Implications for security teams

OpenAI plans to continue advancing its models, applying them to its own defenses and expanding access for trusted defenders. For businesses, the practical implication is to treat independent research, engineering remediation and AI-assisted testing as connected parts of a continuous security process.