Follow us on Google News
OpenAI is planning a security sprint with outside researchers to find vulnerabilities and test fixes across its systems.
What Is OpenAI’s New Security Sprint?
OpenAI has announced plans for an internal security initiative that will bring external researchers together with its engineering teams to identify vulnerabilities, test corrective measures and strengthen its platforms using artificial intelligence.
Announced on November 2, the initiative includes a partnership with Cathedral Systems and expanded financial incentives for independent security researchers reporting serious vulnerabilities. The programme forms part of OpenAI’s efforts to make vulnerability detection and remediation a continuous process rather than a one-time security assessment.
Researchers from Cathedral Systems will work alongside OpenAI employees, using the company’s AI models to examine existing protections and develop additional approaches to security testing. Gavin Kliger, Cathedral Systems’ chief executive and co-founder, said the collaboration would combine his team’s research capabilities with OpenAI’s models and security expertise.
The company has not announced a starting date, specific targets or expected vulnerability findings.
https://the420.in/fcrf-certified-chief-ai-officer-second-cohort-november-2026/
How Will AI Support Vulnerability Testing?
The initiative builds on an earlier security sprint during which OpenAI assigned 25% of its product engineers to security-related work. Engineers used Codex to investigate vulnerability reports, reproduce suspected weaknesses in isolated environments and prepare patches for review. OpenAI’s published Defense Factory architecture describes how AI agents can map systems, investigate potential vulnerabilities, identify responsible teams and prepare tested fixes.
The process uses isolated environments to reproduce software configurations without conducting experiments directly on production systems. Earlier testing also identified operational difficulties, including duplicate vulnerability reports, unclear severity ratings and missing dependencies.
OpenAI found that applying a software patch did not necessarily mean the correction had reached every deployed system, reinforcing the need for independent verification and human oversight.
How Much Can Security Researchers Earn?
Alongside the initiative, OpenAI has increased rewards under its Security Bug Bounty programme. Researchers identifying critical vulnerabilities in core products, including ChatGPT, Codex and the API Platform, can now receive up to $30,000 (approx. Rs. 28 lakhs).
Exceptional critical findings remain eligible for a base award of $100,000, with additional bonuses potentially increasing the total payout to approximately $172,000 (approx. Rs. 1.6 crores). The final reward depends on factors including report quality, researcher loyalty and priority campaigns.
The company distinguishes standard critical vulnerability rewards from exceptional awards, meaning the highest advertised payment is not available for every critical finding.
What Other Security Measures Has OpenAI Announced?
OpenAI has also connected its security efforts to Daybreak for Frontline Defenders, a separate initiative involving a $1 billion commitment over six months. The programme provides subsidised access, training, support and partnerships to essential services and organisations with limited cybersecurity resources. Its objective is to help defenders identify confirmed weaknesses and turn those findings into tested security improvements.
The initiative follows the same approach as the planned security sprint, combining AI-assisted vulnerability research with practical remediation. OpenAI has maintained that human review and independent checks remain necessary, particularly when security decisions could have significant consequences.
Together, the programmes reflect an emphasis on continuous vulnerability assessment, external research collaboration and verification of security fixes before deployment.
