White House Requires OpenAI to Restrict GPT-5.6 to Government Approved Partners

OpenAI confirmed Friday it is restricting GPT-5.6 to a "small group of trusted partners" preapproved by the Trump administration, marking the first time the US government is directly controlling access to a frontier AI model customer by customer.

The company announced three new models, GPT-5.6 Sol, Terra, and Luna, tiered by capability, but said it would only release them to partners vetted by the White House. CEO Sam Altman told staff in a memo that the government would be "approving access customer by customer during this preview period," The Information reported, with a broader rollout expected "a couple of weeks later" if the review goes well.

"We don't believe this kind of government access process should become the long-term default," OpenAI wrote in a blog post. "It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them." The request came after talks with the Office of the National Cyber Director and the Office of Science and Technology Policy, Reuters reported.

Commerce Secretary Howard Lutnick personally called Altman to demand approvals from multiple agencies, according to The Information. The move follows the same playbook the administration used against Anthropic. Two weeks ago, the US government ordered Anthropic to pull its Fable 5 and Mythos 5 models offline for all foreign nationals under an emergency export control directive.

Those models remain unavailable, and Anthropic is fighting the order in court.

Investor David Sacks, who co-leads Trump's council of technology and science advisers, blamed Anthropic CEO Dario Amodei for triggering the crackdown. "Dario came to Washington a few months ago, back in April, and basically said that he had created a cyber weapon called Mythos," Sacks said on a podcast. "And he spiked the cortisol level, got everyone really worried."

OpenAI is not happy about the arrangement but views it as temporary. The company said it is working with the administration to build a "repeatable process for future model releases."

No such voluntary framework exists yet, Trump's AI executive order signed earlier this month called for one, but it has not been implemented. As a result, frontier AI labs are operating in an uncodified interim where government approval is effectively mandatory.

OpenAI plans to expand access next week to include some international partners. The company said it sends the US government a list of proposed customers and receives feedback, but cannot disclose the approval criteria.

The new Sol model is OpenAI's strongest for cybersecurity, coding, and biology benchmarks. The company said it is better at helping users fix vulnerabilities than carrying out attacks and does not cross its internal "critical" risk threshold.Roughly 20 customers have been approved so far.

Stanford cybersecurity expert Alex Stamos criticized the government's approach, saying the national security concerns lack factual basis. "If the administration is honest about wanting the United States to beat China in this race, then this is about the dumbest thing they could possibly do," he told reporters.