OpenAI partners with Hacktron to enhance security testing of AI systems

OpenAI has established a new collaboration with Hacktron to proactively search for security vulnerabilities within its systems. This partnership involves OpenAI engineers and Hacktron security researchers jointly examining high-priority areas, with a specific focus on how the company's systems interface with external third-party services. Furthermore, the agreement extends Hacktron's access to OpenAI's frontier models via Daybreak Red. This initiative aligns with OpenAI's broader commitment to cyber defense, which also includes an upcoming security sprint where Cathedral's researchers will collaborate with internal teams. To encourage external scrutiny, the company has also increased its bug bounty rewards for independent researchers. OpenAI previously engaged with Hacktron through its Security Bug Bounty program and was impressed by the disclosures made by the firm's team, which specializes in building AI agents for penetration testing and red teaming. The collaboration aims to deepen testing of complex service connections, leveraging Hacktron's expertise in uncovering weaknesses in application, cloud infrastructure, and identity systems. OpenAI expressed gratitude to Hacktron and other independent researchers for their contributions to strengthening user security.