Only select partners will initially be able to use the model, while the U.S. government will approve access for each customer individually, the outlet reported.
CEO Sam Altman told employees during a meeting this week that the government would approve access on a case-by-case basis during the pre-release phase, TechCrunch reported.
OpenAI hopes to make the model available to a broader group of users within several weeks if that stage is successful, the outlet reported.
The arrangement would effectively mirror an earlier voluntary decision by Anthropic not to provide general access to its most powerful model.
OpenAI employees also worked closely with the U.S. government while preparing the launch, TechCrunch reported. The limited release was reportedly proposed by the Office of the National Cyber Director and the White House Office of Science and Technology Policy.
The Trump administration initially said it would take a limited role in regulating artificial intelligence. Federal oversight of new models has increased in recent months, however.
Trump signed an executive order in early June recommending that companies voluntarily submit new models to the government for testing and assessment before their public release, TechCrunch reported.
Anthropic previously sparked debate when it announced that its new cyber model, Claude Mythos, would be available only to a limited group of partners through its Project Glasswing program.
The company said the model was too powerful to release widely because misuse could cause more harm than benefit. Experts have debated whether such claims represent a genuine security measure or a marketing strategy.
Cybercriminals have long used automated tools, but generative AI has significantly expanded their capabilities. Large language models can already generate malicious software, while some models may be able to carry out ransomware attacks autonomously.
The main concern surrounding advanced cyber models is that they could identify and exploit software vulnerabilities faster than humans. Hidden flaws in software can provide access to corporate networks, posing a potential risk to organizations with complex IT infrastructure.
The closed nature of access to such models makes their real-world risks difficult to assess.