OpenAI Joins Forces With Cathedral to Fortify AI Cyber Defence
OpenAI teams up with Cathedral and boosts bug bounty rewards to strengthen AI-powered cyber defence against evolving threats.
In a strategic move to stay ahead of evolving digital threats, OpenAI has announced a partnership with the cybersecurity research firm Cathedral. This collaboration aims to strengthen OpenAI cyber defence systems by merging the company's cutting-edge AI models with the specialized expertise of external security researchers.
A Collaborative War Room Against Vulnerabilities
Under this new alliance, researchers from Cathedral will work hand-in-hand with OpenAI's internal teams. The primary goal is to turn the AI's own power against potential threats; the teams will utilize OpenAI's advanced models to proactively hunt for security weaknesses, stress-test existing protections, and engineer robust solutions to mitigate potential risks.
This partnership is set to be a cornerstone of OpenAI's upcoming internal security sprint. Gavin Kliger, the CEO and co-founder of Cathedral, emphasized that OpenAI is setting a vital industry precedent by inviting external scrutiny and using its own technology to continuously harden its defenses.
"OpenAI is setting an important example by partnering with outside security firms and using its own models to continuously test and strengthen its defenses," Kliger stated, adding that the move establishes a continuous, rather than reactive, approach to cybersecurity.
Boosting the Bounty: Incentivizing Security Researchers
Beyond the partnership with Cathedral, OpenAI is throwing more resources into its Security Bug Bounty programme. To attract the world's best "ethical hackers" and security analysts, the company has significantly increased the rewards for those who identify critical vulnerabilities—hiking the payouts by more than four times.
Building on a Legacy of Security Sprints
This is not the first time OpenAI has prioritized a "security-first" internal culture. The company previously conducted a security sprint where a staggering 25 per cent of its product engineers were diverted from their usual tasks to focus solely on finding and fixing system vulnerabilities.
During that initiative, engineers leveraged Codex to:
- Investigate reported security findings.
- Reproduce vulnerabilities within isolated, safe environments.
- Develop and prepare patches for rigorous review.
The Urgency: AI Agents and Emerging Threats
The push for enhanced security comes at a critical time. There are mounting concerns regarding the ability of advanced AI agents to bypass restricted environments, suggesting that as AI becomes more capable, the risks of "jailbreaking" or unauthorized access increase.
OpenAI has acknowledged that traditional cybersecurity measures are no longer enough to combat these emerging threats. Consequently, the company has expanded its monitoring of advanced models across several stages, including training, evaluations, internal usage, and specific external deployments.
The Daybreak Initiative: Supporting Global Infrastructure
In a broader effort to protect the digital ecosystem, OpenAI is launching the Daybreak initiative. Through this program, the company is granting trusted security professionals access to its advanced cybersecurity models.
To ensure that essential services and critical digital infrastructure remain protected, OpenAI has committed USD 1 billion in subsidized access over a six-month period. The company noted that keeping the "defender's window" open requires immediate, industry-wide action to ensure that those protecting the web have the best tools available.