OpenAI has announced an expansion of its Daybreak program, introducing two access tiers and a new model, GPT-5.6-Cyber. This initiative aims to provide frontier intelligence to trusted cybersecurity defenders, preparing them for the increasing use of AI in cyberattacks.
The GPT-5.6-Cyber model is built upon GPT-5.6 Sol and is specifically trained to improve performance on specialized cybersecurity tasks, such as finding zero-day vulnerabilities and developing exploit chains. It also reduces refusals for certain higher-risk, dual-use cyber tasks.
Key Points
- OpenAI is expanding its Daybreak program with two access tiers for approved defenders.
- GPT-5.6-Cyber is a new model available through Daybreak Red access, built on GPT-5.6 Sol.
- GPT-5.6-Cyber is trained to improve capabilities in specialized cybersecurity tasks like finding zero-day vulnerabilities and developing exploit chains.
- The model also reduces refusals for certain higher-risk, dual-use cyber tasks.
- Daybreak Blue access removes system-level guardrails from GPT-5.6 Sol, aiding legitimate defensive work.
- An internal evaluation, Advanced Cybersecurity Completion Rate, shows GPT-5.6-Cyber completes 95.0% of requests, compared to 1.5% for GPT-5.6 Sol and 2.0% with Daybreak Blue access.
- GPT-5.6-Cyber outperforms GPT-5.5-Cyber, which completed 57.3% of requests in the same evaluation.
Context
According to OpenAI, the cybersecurity landscape is evolving rapidly, with threat actors increasingly leveraging AI for cyberattacks. The Daybreak program is designed to equip defenders with advanced AI capabilities before offensive AI tools become widespread. The program's tiers, Daybreak Blue and Daybreak Red, offer different levels of access and model capabilities to suit various defensive needs.
Why It Matters
This development provides cybersecurity professionals with more specialized and less restrictive AI tools for defensive operations. The reduced refusal rates and enhanced capabilities of GPT-5.6-Cyber could allow defenders to accelerate vulnerability research, incident response, and security assessments, potentially shifting the advantage in the ongoing cyber arms race.
What To Do
- Note the distinction between Daybreak Blue and Daybreak Red access tiers for GPT-5.6 Sol and GPT-5.6-Cyber.
- Compare the refusal rates of GPT-5.6-Cyber (95.0%) with GPT-5.6 Sol (1.5%) for advanced cybersecurity tasks.
- Watch for further details on how to gain approved access to the Daybreak program and its models.
- Consider the implications of specialized models like GPT-5.6-Cyber for vulnerability discovery and exploit development workflows.