OpenAI has made significant strides in cybersecurity by announcing a series of initiatives aimed at enhancing vulnerability detection and support for open-source projects. This includes the release of an updated version of its AI model, GPT-5.5-Cyber, designed to improve the identification and remediation of software vulnerabilities.
Release of GPT-5.5-Cyber
On June 23, 2026, OpenAI introduced the enhanced GPT-5.5-Cyber, which is touted as its most advanced model for finding and fixing bugs in code. The company claims that this model can conduct deeper analyses across extensive codebases, identifying security-relevant components and validating potential issues in controlled environments. It is also capable of developing and testing patches, as well as preparing evidence for human review.
Performance Benchmarks
OpenAI has reported that the updated GPT-5.5-Cyber has outperformed its predecessor in several benchmarks. In the CyberGym test, it achieved a success rate of 85.6%, compared to 81.8% for the earlier version. In the ExploitGym evaluation, the new model scored 39.5%, while the previous version managed only 25.95%. Additionally, in the SEC-bench Pro test, GPT-5.5-Cyber reached 69.8%, surpassing the earlier model’s 63.1%.
Expansion of Partner Program
OpenAI has also expanded its OpenAI Daybreak Cyber Partner Program, which now includes around 30 security vendors and service providers. Only these selected partners will have access to the updated GPT-5.5-Cyber model, with plans to onboard more organizations in the future.
Patch the Planet Initiative
Another notable announcement is the launch of the “Patch the Planet” initiative, co-founded with Trail of Bits. This program aims to assist open-source project maintainers in identifying and addressing vulnerabilities. Participants will receive access to ChatGPT Pro, the Codex Security scanner, and API credits for development and automation. In its initial week, the initiative reportedly uncovered hundreds of bugs and generated numerous pull requests across various projects, including cURL and Python.
OpenAI’s Codex Security plugin, which enables automated security workflows, has also been released. This plugin has scanned over 30 million commits and identified a substantial number of vulnerabilities, with many marked as fixed by human reviewers and AI.
This article was produced by NeonPulse.today using human and AI-assisted editorial processes, based on publicly available information. Content may be edited for clarity and style.
