OpenAI expands Daybreak cybersecurity initiative as AI agent threats evolve

OpenAI CEO Sam Altman arrives for a meeting at the White House in Washington, D.C., U.S., July 30, 2026. Nathan Howard | ReutersOpenAI on Monday said it is expanding Daybreak, its exclusive cybersecurity initiative, to help give its participants access to the "right capabilities" that they need to help defend themselves from attackers. The company first introduced Daybreak in May, shortly after its chief rival Anthropic captivated Wall Street and the U.S. government by launching its own cybersecurity coalition called Project Glasswing. OpenAI positioned Daybreak as a way for its ecosystem partners to use its most advanced artificial intelligence models to adapt to a rapidly changing threat landscape. OpenAI said Monday that it is expanding the program to include two different access tiers: Daybreak Blue and Daybreak Red. The expansion follows a number of cybersecurity incidents that AI developers —including OpenAI, Anthropic and Meta — have disclosed in recent weeks. In each case, an AI model accessed systems that should have been off limits as part of cybersecurity testing, prompting industry researchers and government officials to call for stronger protections. "As the threat landscape evolves, we're putting frontier intelligence in the hands of trusted defenders before attackers can deploy offensive AI at scale," OpenAI said in a post on X on Monday.Daybreak Blue will give users unique access to OpenAI's advanced general-purpose models, as the safeguards will be altered to allow for defensive security work, the company said in a blog post. Daybreak Red participants will be able to go a step further, leveraging OpenAI's "purpose-trained cybersecurity models" for security testing, vulnerability research and exploit validation. Read more CNBC tech newsHugging Face hack marks start of dangerous AI cyber era and many firms 'don't even know it'How a small Israeli startup was linked to rogue AI hacks at OpenAI, Anthropic and MetaWorld's biggest chipmaker TSMC's sales surge 45% amid buoyant AI demandMeta to open source its most powerful AI model as it takes swipe at OpenAI, AnthropicOpenAI is also launching a new AI model, GPT‑5.6‑Cyber, which will be available to Daybreak Red users. The company said the model is built on its most powerful publicly available offering, GPT-5.6 Sol, but it's designed to improve capabilities and reduce refusals on certain specialized cybersecurity tasks. The company said it recommends Daybreak Blue as the starting point for most organizations. OpenAI has repeatedly warned about AI models' advancing cyber capabilities in recent months. The company said last week that it is pausing some internal activities involving an upcoming model called Astra, which has demonstrated "significant advancements in agentic coding and cybersecurity" during testing. It said it is working to assess these capabilities and implement more robust safeguards and security controls. "We're committed to working alongside governments, safety institutes, and civil society to ensure that the frontier capabilities of models like Astra, and those that follow, are deployed responsibly and broadly for the benefit of all humanity," OpenAI said. WATCH: OpenAI, Anthropic agents participate in new ‘unsanctioned’ AI behavior