OpenAI deploys GPT-5.6-Cyber to fight AI agent hacks
openai expanded its daybreak cyber defense service with a new tier red and a specialized model gpt-5.6-cyber, available only to partners like accenture and ibm. the release highlights the growing role of frontier ai models in active cyber warfare, as attackers use similar agents to automate breaches.
openai launched a significant expansion of daybreak, its cybersecurity offering, splitting the service into blue and red tiers to differentiating between basic incident response and advanced offensive security research. the red tier introduces gpt-5.6-cyber, a model derived from gpt-5.6 sol, which is currently restricted to select enterprise partners including accenture, ibm, and crowdstrike. the company is releasing these tools to help defenders prepare for a new wave of autonomous ai agents that are increasingly being used by threat actors to compromise systems and engineer attacks. by providing access to frontier models, openai aims to close the gap between attackers using ai for automation and defenders relying on traditional methods, acknowledging that the window for preparation is narrowing.