Hackers are already using AI to move faster than ever, and OpenAI just answered back. The company this week rolled out a new OpenAI cyber model built specifically for cybersecurity defense, expanding its Daybreak program right as AI-powered attacks continue to multiply.
Inside OpenAI’s Expanded Daybreak Program
Daybreak isn’t brand new — OpenAI launched the cyber defense service earlier this year, bundling access to models, tools, and workflows built for security teams. What’s changed is the scale and structure of what defenders can now access. TechCrunch
OpenAI said the program now splits into two tiers: Blue and Red. Both give approved customers access to the company’s limited-access frontier cyber models, though each tier serves a different kind of user. TechCrunch
Blue is described as the “recommended starting point for most defenders,” suggesting it covers most enterprise needs without extra complexity. Red, meanwhile, hands users a broader and more powerful toolkit designed for advanced security work. TechCrunch
Meet GPT-5.6-Cyber
The headline release here is GPT-5.6-Cyber, a purpose-built model available only through the Red tier. It’s built on top of GPT-5.6 Sol and adds enhanced capabilities for specialized cybersecurity tasks, according to OpenAI. TechCrunch
Right now, access is tightly controlled. The model is only available to “trusted customer partners,” reportedly including Accenture, IBM, CrowdStrike, and Cloudflare. That short list reflects just how cautiously OpenAI is treating a model capable of vulnerability research and security testing. TechCrunch
Why This Launch Matters Right Now
This isn’t happening in a vacuum. Cybersecurity defenders have been running into high refusal rates across frontier AI models, as labs try to balance giving legitimate security teams useful tools without leaking those same capabilities to attackers. Axios
Notably, GPT-5.6-Cyber only reached the “High” cyber capability threshold under OpenAI’s Preparedness Framework, a more measured classification than some of the company’s earlier, more alarming warnings about future models. Still, the timing lines up with a summer of real incidents involving AI-driven intrusions across the industry, pushing every major lab to treat AI cybersecurity threats as an urgent priority rather than a future problem. Axios
Beyond OpenAI, the broader industry is racing to match pace. Competing labs have rolled out their own cyber-focused models in recent months, turning AI-assisted defense into a genuine arms race between attackers and defenders.
The Bottom Line
As AI-led cyberattacks grow more sophisticated, the tools built to stop them are evolving just as fast. OpenAI’s expanded Daybreak program and the new GPT-5.6-Cyber model show that defense is becoming as automated and AI-driven as the threats themselves.
For security teams watching this space, the message is clear: staying ahead now means adopting AI-powered defense before attackers gain the upper hand. Keep an eye on how access to these frontier cyber models expands in the months ahead.
