GoKawiilOpenAI's GPT-6 Astra found two unknown zero-day exploits during security testing
OpenAI tested its new Astra model on new benchmarks measuring exploit development and binary reverse engineering, finding it far outperformed the prior GPT-5.6 Sol model. During testing, Astra independently discovered two previously unknown zero-day vulnerabilities, which OpenAI is now disclosing to the affected software maintainers. Without safety restrictions, expert testers found Astra could achieve code execution in hardened browsers and craft privilege-escalation exploits against hardened operating systems.
GoKawiil's interpretation of the reporting above, not reported fact.
The jump in autonomous exploit-finding and reverse-engineering ability shows AI models are approaching a threshold where they can meaningfully accelerate both offensive hacking and defensive security work. OpenAI's decision to keep safeguards on advanced exploit creation, while planning to loosen them for defensive uses, reflects how seriously it's treating the dual-use risk of increasingly capable cyber-focused models.
- Astra outperformed GPT-5.6 Sol on new exploit-development and reverse-engineering benchmarks.
- The model discovered two real zero-day vulnerabilities during testing, now being responsibly disclosed.
- OpenAI currently blocks Astra from creating proof-of-concept exploits but plans to ease restrictions for defensive security tasks.
No Starch Press "Practical Malware Analysis — If reverse-engineering benchmarks like SRE-Bench caught your eye, this is the classic hands-on guide to pulling apart binaries and understanding what code really does without source. It walks through disassembly, debugging, and behavioral analysis with real labs, so you can follow along with the same skills these models are being measured on.
See No Starch Press "Practical Malware Analysis on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: openai.com, 2026-09-03
Published there as: “GPT-6 Astra”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.