OpenAI said Friday it is limiting access to its new GPT-5.6 artificial-intelligence models after discussions with the Trump administration, the latest sign that the White House is tightening control over powerful AI systems amid national security concerns.
The company will initially make GPT-5.6 available to a small group of customers approved by the administration, according to a person familiar with the matter reported by The Wall Street Journal. Three models operating under the GPT-5.6 umbrella will be available, tailored to reasoning, everyday tasks, and more intense workloads, with different costs and speeds based on complexity.
OpenAI said in a blog post announcing the models that it hopes to make GPT-5.6 generally available in the coming weeks and described the government’s current approval process as a transition period while President Trump’s recent executive order on model oversight is implemented.
“We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders and global partners who need them,” the company said, adding that it has extensive security safeguards in place. “We are taking this short-term step because we believe it is the strongest path to broader availability.”
The discussions about GPT-5.6 have coincided with a Trump administration ban on foreigners accessing two of Anthropic’s newest models — Fable 5 and Mythos 5 — which led Anthropic to halt all access to comply. Anthropic had previously worked with the administration on a limited rollout of an earlier version of Mythos. The latest Anthropic models have been shut down for two weeks while both sides discuss how to address security concerns, fueling criticism from AI analysts who say the White House is picking winners and losers in the industry.
OpenAI followed a similar process with versions of GPT-5.5, which demonstrated an ability to discover vulnerabilities in software that could be used in cyberattacks, according to the Journal. The similar capabilities of Anthropic’s Mythos models prompted the White House to increase its oversight of the industry.
OpenAI and the administration have been discussing the GPT-5.6 rollout since before the ban on Anthropic’s models, people familiar with the matter said. CEO Sam Altman and Commerce Secretary Howard Lutnick discussed the phased rollout on Wednesday, the people said. The Commerce Department is one of the main agencies overseeing AI models and houses the Center for AI Standards and Innovation, a government testing unit that reviews models before they are released. CAISI has been reviewing GPT-5.6 just as it tests other models, the people said.
The recent executive order gave cybersecurity and national-security officials a bigger role in model evaluation, worrying OpenAI and other companies that have been pushing for CAISI to maintain a significant presence.
Trump, Altman, and other world leaders and AI executives recently discussed model oversight and safeguards during a Group of Seven meeting in France.
GPT-5.6 has demonstrated improved performance in completing tasks autonomously, a key gauge of model performance, as well as enhanced knowledge about biology and cybersecurity, the company said. AI developers are working with the government and other businesses to use these capabilities to bolster security before wider release.
OpenAI has backed Trump’s approach to keeping the U.S. ahead of China in the AI race and accelerating the build-out of data centers needed to train AI models. Co-founder and President Greg Brockman has donated to a Trump-aligned political-action committee.