OpenAI Partners with Hacktron to Enhance System Security

Partnership Overview

OpenAI is collaborating with Hacktron, a security research firm specializing in AI-driven code review, penetration testing, and red teaming, to identify and remediate vulnerabilities within OpenAI's infrastructure. This partnership prioritizes testing how OpenAI's systems interact with third-party services, leveraging Hacktron's expertise in uncovering flaws across applications, cloud infrastructure, and identity systems.

Expanded Access to Frontier Models

As part of this initiative, OpenAI is granting Hacktron researchers expanded access to its frontier models through the Daybreak Red program. This access is intended to facilitate deeper security testing and research into the vulnerabilities that may arise as AI agents increasingly connect disparate services.

Broader Security Strategy

This collaboration is part of a wider effort by OpenAI to accelerate cyber defense and harden its systems. Recent initiatives include:

• Security Sprints: A planned collaboration with Cathedral researchers to conduct intensive security testing.
• Increased Bug Bounty Rewards: OpenAI has raised its bug bounty payouts to incentivize independent researchers to identify vulnerabilities that internal testing may have missed.

Rationale for Collaboration

OpenAI selected Hacktron based on the firm's performance within the existing Security Bug Bounty program. OpenAI noted that Hacktron's ability to build AI agents for security tasks aligns with the company's goal of using AI to move faster and address more complex security challenges. The partnership aims to strengthen defenses by testing the security of connections between AI agents and external services.