OpenAI partners with Cathedral to strengthen AI cyber defenses

A new defensive partnership

OpenAI has entered into a strategic partnership with the cybersecurity research firm Cathedral. This collaboration aims to bolster AI-powered cyber defense systems by combining internal technology with external expertise. The goal is to create a more robust framework for identifying and mitigating digital threats.

Under the terms of this agreement, researchers from Cathedral will work directly alongside OpenAI teams. They will use OpenAI's own advanced models to probe for security weaknesses and test the effectiveness of current safeguards. This collaborative approach is designed to create a continuous cycle of testing and improvement for the company's infrastructure.

Gavin Kliger, the CEO and co-founder of Cathedral, noted that the partnership combines AI capabilities with external cybersecurity knowledge. He stated that OpenAI is setting an example by using its own models to continuously test and strengthen its defenses. This move highlights a growing trend of using the same technology that creates risks to also provide solutions.

Expanding the bug bounty program

To further encourage independent researchers to find flaws, OpenAI is significantly expanding its Security Bug Bounty program. The company is increasing the rewards for identifying critical vulnerabilities by more than four times. This financial incentive is intended to draw more eyes to the systems to find and report security gaps before they can be exploited.

The initiative is part of a broader effort to proactively manage the risks associated with large-scale AI deployments. By offering higher payouts, OpenAI hopes to foster a community of researchers who can act as an external layer of defense. This strategy allows the company to identify issues that internal teams might miss during standard development cycles.

The move comes as concerns grow regarding the ability of advanced AI agents to bypass restricted environments. As these models become more sophisticated, the potential for them to find unintended paths into secure systems increases. Strengthening the bug bounty program is a direct response to the need for more rigorous, constant testing of these evolving capabilities.

Lessons from previous security sprints

This new partnership builds on a previous security sprint conducted by OpenAI. During that period, the company dedicated 25 percent of its product engineers specifically to identifying and fixing vulnerabilities. This large-scale allocation of human resources showed a commitment to prioritizing security over rapid feature expansion.

During that previous sprint, engineers utilized Codex to investigate various security findings. They used the model to reproduce vulnerabilities within isolated environments, which allowed them to understand the root cause of a problem without risking the live system. Once a vulnerability was reproduced, the team could then prepare specific patches for review.

OpenAI has noted that traditional cybersecurity measures alone are no longer sufficient to handle emerging threats. The company is now moving toward a model of continuous testing and improvement. This shift is necessary because the nature of AI-driven threats requires a more dynamic and integrated security posture than static firewalls can provide.

The Daybreak Initiative

In a separate move to support the broader tech ecosystem, OpenAI is extending access to its advanced cybersecurity models through the Daybreak initiative. This program is designed to provide trusted security professionals with the tools they need to protect essential services. It aims to level the playing field for defenders who may not have the resources of larger corporations.

OpenAI has committed USD 1 billion in subsidized access to these models over a period of six months. This funding is specifically intended to support organizations that protect digital infrastructure and essential services. By lowering the cost of high-end security tools, OpenAI hopes to improve the overall resilience of the internet.

The company stated that keeping the defender's window open requires action across the entire industry. They believe that the work to secure digital infrastructure needs to start immediately. The Daybreak initiative represents a significant investment in the security tools that will be used to combat future cyber threats.

Monitoring and future risks

OpenAI is also increasing the scope of its internal monitoring for its advanced models. This monitoring now covers training, evaluations, internal use, and selected external deployments. By watching how models behave in various stages of their lifecycle, the company can identify potential security risks earlier in the process.

The expansion of monitoring is a response to the increasing complexity of AI systems. As models are deployed in more diverse environments, the surface area for potential attacks grows larger. Continuous oversight ensures that the company can react quickly if a model begins to behave in an unexpected or insecure manner.

Looking forward, the partnership with Cathedral and the Daybreak initiative suggest a long-term strategy for AI safety. OpenAI is positioning itself as a leader in the development of defensive AI. As the technology continues to evolve, the focus will likely remain on creating systems that can withstand increasingly sophisticated automated attacks.