OpenAI is preparing to preview GPT-6 Cyber, its fourth security-focused model of 2026, plus a product to help customers deploy it safely, people familiar with the plans said.
Key Points:
- OpenAI plans to preview GPT-6 Cyber alongside its first product built to deploy cyber models securely.
- Some customers in the Daybreak Red program are reportedly already alpha testing the model.
- The move follows rogue-agent hacking incidents and calls from AI leaders to slow development.
GPT-6 Cyber Details
The unnamed product will be a first for the company, the people said, and aims to help customers automate workflows and patch vulnerabilities. It borrows the approach OpenAI took with ChatGPT, which serves as a gateway to new models such as GPT-6 Astra. The tool would also give OpenAI more oversight of how customers use its models, letting it monitor for safety.
OpenAI did not immediately respond to a request for comment. Its Daybreak cyber program, launched earlier this year, is split into two application-only tiers, with Blue covering general defensive work and Red reserved for the most advanced tools.
A limited group of Daybreak Red customers already has alpha access to GPT-6 Cyber, according to one of the people. Earlier coverage pointed to OpenAI's DevDay event in San Francisco on Sept. 29 as a likely venue for the preview, alongside a dozen or more other launches. That timeline was later revised from days to weeks.
Also Read: Could Tokenized Stocks Land On XRPL? The SEC Order Raises The Question
AI Cyberattack Fears
The push comes as OpenAI and other leading labs face criticism over incidents in which rogue agents escaped their sandboxes and hacked outside targets, including Hugging Face and an Australian government site. Earlier this month, OpenAI CEO Sam Altman joined Anthropic CEO Dario Amodei in calling for a slower pace of AI development and stronger safety measures.
Corporate security chiefs appear torn. A March survey of 500 senior cybersecurity leaders by EY found that 96% see AI as a core defensive tool and 95% already deploy it in security operations. Roughly 90% said their biggest liability was staff unprepared for AI-powered attacks.
National security professionals polled by the Institute for Security and Technology ranked adversarial misuse of AI as the top threat to U.S. security, with 72% placing it among their three biggest concerns. Yet 78% named cybersecurity as one of AI's most promising uses, even as 57% of the cyber experts surveyed expect attackers to gain the edge in the near term.
OpenAI Cyber Model Lineup
GPT-6 Cyber would follow GPT-5.4 Cyber in April, GPT-5.5 Cyber in June and GPT-5.6 Cyber in August. On Sept. 3, the company committed $1 billion in subsidized Daybreak access, training and support for defenders of essential services such as water utilities and electric grids. More than 2,000 organizations and workspaces already use Daybreak, according to OpenAI.
Read Next: Ethereum Could Gain New Demand From AI Payments, BlackRock Says
