OpenAI says GPT-5.6-Cyber finds hundreds of flaws before launch
This digest was compiled by AI from multiple sources — links to the originals are below.
OpenAI says its GPT-5.6-Cyber model has found hundreds of serious software vulnerabilities, including zero-days, before general release. The model completed 95% of advanced cybersecurity tasks in internal testing, up from 1.5% for GPT-5.6 Sol. OpenAI expanded its Daybreak program with Blue and Red tiers for approved defenders.
GPT-5.6-Cyber Performance
OpenAI said GPT-5.6-Cyber completed 95% of requests for advanced cybersecurity tasks in its Advanced Cybersecurity Completion Rate test, compared with 1.5% for GPT-5.6 Sol with standard safeguards. The model finds zero-day vulnerabilities, develops exploit chains, validates fixes, and supports high-risk defensive work. One partner said the model completed work in less than a day that earlier models had failed to resolve after weeks.
Chrome V8 Zero-Day Findings
OpenAI said GPT-5.6-Cyber uncovered two previously unknown vulnerabilities in Chrome's V8 JavaScript engine before launch. Researchers validated the findings and reported them to Google, which fixed the issue as CVE-2026-15903. The model also identified at least five vulnerabilities in a popular mobile operating system, plus flaws in a database and a kernel.
Daybreak Program Expansion
OpenAI expanded its Daybreak cybersecurity initiative with Blue and Red tiers for approved defenders doing advanced work. Blue provides frontier general-purpose models for vulnerability discovery, malware analysis, incident response, and secure code review. Red offers specialized purpose-built cyber models for vulnerability research, exploit validation, and security testing. The company argued that threat actors will increasingly use AI at speed, giving defenders a shrinking response window.
What's Next
OpenAI is expected to make GPT-5.6-Cyber more broadly available through Daybreak, though no release date was given. It remains unclear whether organizations can patch the growing flow of AI-discovered vulnerabilities faster than attackers can exploit them.
1 source
OpenAI says GPT-5.6-Cyber finds hundreds of flaws before launch
