Artificial intelligence is rapidly changing the cybersecurity battlefield, and OpenAI is responding with a new model designed to help defenders fight back. The company has expanded its Daybreak cyber defense service with a new security-focused model, GPT-5.6-Cyber, as concerns grow over AI agents being used to carry out increasingly sophisticated cyberattacks.
The move comes as reports of AI systems being used in real-world hacking and social engineering incidents continue to emerge. From compromised platforms to attempts to manipulate people into granting access, AI agents are showing that cyber threats can now move faster and operate on a much larger scale. OpenAI says its expanded Daybreak service is intended to give security teams the tools they need to stay ahead of these threats.
Daybreak will now be divided into two tiers, known as Blue and Red. The Blue tier is aimed at most organizations and provides tools for tasks such as incident response, malware analysis and checking whether security patches have been properly applied. The Red tier goes further, giving approved users access to specialized cybersecurity models for security testing and vulnerability research.
At the center of the Red tier is GPT-5.6-Cyber, which OpenAI says is built on GPT-5.6 Sol and enhanced for specialized cybersecurity tasks. However, access is currently restricted to trusted customer partners, reportedly including major technology and cybersecurity companies such as Accenture, IBM, CrowdStrike and Cloudflare. The limited rollout reflects the sensitive nature of powerful AI models that can potentially be used for both defensive and offensive cybersecurity work.
OpenAI believes the cybersecurity industry has a shrinking window to prepare for AI-powered attacks before they become faster, more autonomous and harder to stop. While some critics argue that AI companies also benefit commercially from promoting the dangers their own technology can create, businesses are increasingly looking to these same companies for protection. With AI becoming both a potential weapon and a defensive tool, the cybersecurity race may ultimately come down to which side can adapt faster.
source: techcrunch
