OpenAI Launches ChatGPT Cyber Model as AI-Driven Vulnerability Discovery Outpaces Defenders
A startup CTO used OpenAI's GPT 5.6 Sol, accessed through the Daybreak Blue program, to find a severe Ubuntu server vulnerability in two hours. OpenAI then released ChatGPT Cyber to top-tier Daybreak Red users, saying it had found 948 potential Linux kernel issues and 81 novel Firefox vulnerabilities, and announced a $1 billion initiative for critical-infrastructure defenders. Critics say defenders are under-resourced relative to attackers, especially those using open-weight models with stripped guardrails.
Skynet Chance (+0.05 pp): Frontier models can now find convoluted, hard-for-humans exploits in hours, and attackers can use open-weight models with removed guardrails. This raises the risk of AI-enabled compromise of critical systems and shows that misuse is outpacing control, though OpenAI's defensive initiatives partly offset this.
Skynet Date (1 day sooner): The article expects 'extremely crazy bugs' to be found and exploited within months, while defense remains under-resourced. That compresses the timeline over which offensive AI capabilities become widely dangerous.
AGI Progress (+0.02 pp): Autonomously finding novel, complex vulnerabilities across large codebases shows strong expert-level reasoning and code understanding. Spending $6 million in tokens on an open fluid dynamics problem also suggests models are being applied to frontier scientific work.
AGI Date (no change): Rapid releases of increasingly capable specialized models, together with large compute allocations to expert tasks, point to continued fast capability gains.
<< All AI news for November 5, 2026
[ Get the daily index digest on Telegram → ]Related AI News
- OpenAI Launches GPT-6 Cyber, an Aligned Agent for Vulnerability Discovery and Patching
- OpenAI Autonomous Pipeline Finds and Patches 948 Linux Kernel Privilege-Escalation Bugs
- Daybreak Red Brings GPT-5.6 Cyber to AWS GovCloud via Amazon Bedrock
- OpenAI Cuts Funding Round Target to $20 Billion Amid Tepid Investor Demand
- Attackers Backdoor Official Anthropic Claude Extension on Cursor Marketplace, Exposing Up to 5,000 Developers