OpenAI has postponed the extensive public rollout of its latest frontier artificial intelligence model, GPT-5.6, after the U.S. government requested an opportunity to examine the technology before it reaches a wider audience. Rather than making the model immediately available to all users, the company will begin with a restricted deployment involving a small number of carefully vetted partners whose identities have been disclosed to federal authorities.
The temporary decision surfaces an increasingly cautious approach toward highly capable AI systems as governments evaluate their potential impact on national security. Policymakers have become more concerned that advanced generative AI models, while offering substantial benefits across research, software development and cybersecurity, could also be exploited to support sophisticated cyberattacks, automate vulnerability discovery, generate convincing phishing campaigns or assist other malicious activities if deployed without adequate safeguards.
According to OpenAI, the limited rollout is intended to provide government officials with an opportunity to study the model's capabilities and assess possible security risks before broader public access is granted. The company said it has already briefed the U.S. government on GPT-5.6 and its expected capabilities and described the current arrangement as an interim measure while it works with Washington to establish a more structured framework for releasing future frontier AI models.
Chief Executive Officer Sam Altman publicly expressed support for rigorous safety evaluations but questioned whether government agencies should determine which organizations receive early access. In a post on X, Altman said extensive testing of advanced AI systems is appropriate, while arguing that customer selection should remain outside government control.
The latest development follows an executive order signed earlier this month by President Donald Trump establishing a voluntary process under which developers of designated "covered frontier models" may provide the U.S. government with access to their systems for up to 30 days before they are released to trusted external partners. The initiative is designed to give officials time to evaluate emerging security concerns and strengthen oversight of increasingly capable AI technologies before wider deployment.
OpenAI stated that restricting access during this initial period represents what it believes is the most practical route toward making GPT-5.6 more broadly available in the coming weeks while discussions continue with the Administration on implementing the cyber-focused executive order and developing a repeatable review process for future launches.
The company added that engineering teams will continue conducting extensive safety evaluations and work closely with early partners throughout the testing phase. At the same time, OpenAI cautioned that the current level of government access should remain a temporary measure rather than becoming a permanent requirement for future AI releases. It also declined to identify the organizations participating in the initial rollout.
OpenAI further warned that prolonged restrictions on access to frontier AI systems could slow innovation across multiple sectors. The company noted that developers, businesses, cybersecurity professionals and international collaborators all rely on access to advanced models to build defensive security tools, strengthen research, develop enterprise applications and accelerate responsible AI adoption.
Leading the new product family is GPT-5.6 Sol, which OpenAI describes as its most capable model to date. The release also includes Terra, positioned as a mid-range model, and Luna, a lower-cost alternative intended to make advanced AI capabilities available at a lower price point across a wider range of use cases.
The government's heightened scrutiny extends beyond OpenAI. Earlier this month, Anthropic was instructed by U.S. authorities to suspend access to its frontier AI models for foreign nationals because of national security concerns. The company continues to face an ongoing legal and regulatory dispute with the government over those restrictions, illustrating the growing debate surrounding oversight of advanced artificial intelligence systems.
The developments come as both OpenAI and Anthropic have confidentially submitted paperwork for U.S. initial public offerings. Separately, The New York Times reported that OpenAI is considering postponing its public market debut until next year.
The developing relationship between AI developers and governments illustrates how the deployment of frontier models is becoming closely linked with cybersecurity and national security policy. While companies continue to pursue increasingly powerful AI capabilities, regulators are placing greater emphasis on evaluating how these systems could influence cyber defense, critical infrastructure protection and the misuse of AI by malicious actors before they are released at scale.
