OpenAI partners with Hacktron to hunt for vulnerabilities in its systems

OpenAI published a notice on its website on November 6 announcing a partnership with security firm Hacktron aimed at proactively finding vulnerabilities in its systems.

The notice lays out three areas of work: finding vulnerabilities, testing connections to third-party services, and strengthening defenses with frontier AI. That scope suggests something broader than a one-off penetration test — it spans exposure review through to building defensive capability.

The third-party connection testing deserves separate attention. Model products typically connect to many external tools, plugins and data sources, and each connection can become a new attack path; having an outside team examine those integrations brings risk at the integration layer into scope.

The notice also mentions using frontier AI to strengthen defenses, placing AI both as the protected object and as a defensive tool — using model capability to find and respond to attacks against model systems themselves.

For a company whose core asset is model capability, bringing in an external security team is a signal that security is moving from after-the-fact response toward routine, externally validated testing.

The notice does not disclose the number of vulnerabilities, the specific testing scope, or a timeline, so what can be confirmed today is the framework and direction of the partnership rather than results already produced.

Three things are worth watching: whether Hacktron publicly describes the classes of vulnerabilities it finds, whether third-party connection testing becomes a reusable process, and what form the frontier AI defenses actually take.