OpenAI Partners With Hacktron to Hunt for Security Flaws in Its AI Systems

OpenAI has announced a partnership with cybersecurity research company Hacktron aimed at finding and fixing vulnerabilities in its systems, with a particular focus on the risks that emerge when its AI models connect with third-party services. Under the arrangement, engineers from OpenAI and security researchers from Hacktron will jointly identify which areas should be tested first.

As part of the deal, OpenAI will widen Hacktron's access to its most advanced AI models through its Daybreak Red program. The company said the collaboration is intended to reinforce its security posture as AI systems grow more capable and increasingly interact with external applications and infrastructure.

The Hacktron partnership follows a series of other cybersecurity moves OpenAI announced this week. The company said researchers from Cathedral will work alongside its internal teams during its next security sprint, and it has also raised the rewards on offer through its Security Bug Bounty program to encourage independent researchers to surface flaws its own testing may have overlooked.

OpenAI said it first became aware of Hacktron through that bug bounty program, where the firm's researchers identified and responsibly disclosed security weaknesses. Hacktron builds AI agents for code review, penetration testing and red teaming, and its research has flagged vulnerabilities tied to the way applications, cloud infrastructure and identity systems interact.

Hacktron CEO Zayne Zhang said the partnership would pair the company's offensive security expertise with OpenAI's advanced models to uncover complex weaknesses across interconnected systems. He added that effective cybersecurity increasingly depends on combining human expertise with AI-powered tools rather than simply growing security teams.

OpenAI said the collaboration will let both teams investigate vulnerabilities more thoroughly and feed their findings back into stronger protections. The effort reflects a broader push across the AI industry to stress-test models and the services they connect to before flaws can be exploited.

The financial terms of the partnership were not disclosed.