OpenAI launched GPT-6 Cyber, a model for finding exploitable vulnerabilities and developing fixes, trained to stay within authorized scope.
Why it matters: A frontier model built to find and fix vulnerabilities raises dual-use stakes, making its scope controls central to its safety case.
OpenAI describes Control Room, a system that tracks model workloads across its research infrastructure and can stop them quickly over safety or security concerns.
Why it matters: Shows a frontier lab building an internal kill mechanism for compute jobs, a concrete safeguard that other labs may be expected to match.
OpenAI introduced a National Community Compact and Compute Campus Policy Blueprint for responsible AI infrastructure.
Why it matters: Offers a company-authored template for community terms on data centres, which governments may use to judge AI infrastructure.
OpenAI describes an autonomous pipeline that found, validated and generated patches for 948 Linux kernel privilege-escalation vulnerabilities.
Why it matters: Shows frontier AI automating vulnerability discovery and patching at scale, a defensive gain that also shows how cheaply exploit-finding can now be done.
Daybreak Red, which brings GPT-5.6 Cyber to approved customers, is now available in AWS GovCloud through Amazon Bedrock.
Why it matters: Puts a cyber-capable GPT model inside US government cloud for authorised users, extending controlled access to dual-use security tools.