OpenAI Unveils GPT-6 Astra, Scores 100% on ExploitBench — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI unveiled GPT-6 Astra, scoring 100% on ExploitBench versus 78.5% for predecessor GPT-5.6 Sol, plus 98% on FrontierMath Tier 4 and 99.9% on ARC-AGI-3, after the model reached the "Critical" cybersecurity threshold under OpenAI's Preparedness Framework
- GPT-6 Astra is rolling out to ChatGPT Plus, Pro, Business, and Enterprise users, and is also available via the OpenAI API, Microsoft Azure, and AWS Bedrock
- OpenAI is releasing a restricted version limited to secure code review and patching that refuses proof-of-concept exploit requests, and plans to ease safeguards via "Daybreak" in the coming weeks to unlock vulnerability validation, malware analysis, and detection engineering
- OpenAI launched "Daybreak for Frontline Defenders," committing $1 billion in subsidized model access, hands-on training, and technical assistance to critical infrastructure sectors including water systems, electricity providers, state and local governments, banks, non-profits, and open-source maintainers
- OpenAI added stronger jailbreak robustness, expanded monitoring, and extra misalignment safeguards to the initial release, alongside a U.S. MS-ISAC pilot giving public-sector and water-utility defenders Daybreak access
Why it matters: OpenAI is shipping a model whose internal variant can already turn known software vulnerabilities into working exploits at a perfect rate, yet paying ChatGPT users only get access to a stripped-down defensive build — a split that defines who touches frontier cyber capability first. The accompanying $1 billion Daybreak program skews that early access toward water utilities, banks, and state/local governments rather than the broader enterprise market.
Ask SkimNews
