Trump restricts OpenAI's GPT-5.6 Sol, second AI model blocked in 30 days

The Trump administration's decision to restrict OpenAI's GPT-5.6 Sol marks the second time in a month the White House has intervened in frontier AI model releases.

The White House ordered OpenAI to limit access to its GPT-5.6 Sol model to government-approved customers, extending a regulatory clampdown that forced Anthropic to pull two models from the market in June.

"Most enterprises are already operating in multi-model, multi-agent environments," said Felix Van de Maele, CEO of data intelligence platform Collibra. "When a company gets 90 minutes to pull a model deployed to hundreds of millions of people because of a competitor's complaint, that need becomes urgent fast."

OpenAI said July 7 that GPT-5.6 Sol, its strongest model with improvements in coding, biology and cybersecurity tasks, would be available only to a select group of trusted partners whose participation has been shared with the government. The move follows the Commerce Department's June 12 order blocking foreign nationals from using Anthropic's Claude Fable 5 and Mythos 5 models after Amazon flagged a method to bypass Fable 5's safeguards. Anthropic was given 90 minutes to comply and took both models offline for more than two weeks before the administration lifted restrictions last week.

The interventions signal a sharp shift from the Trump administration's light-touch approach last year. Trump signed an executive order in June establishing a voluntary framework for government testing of frontier models before release, but the pressure on Anthropic and OpenAI suggests the line between voluntary and mandatory is blurring — raising questions about which companies control access to the most powerful AI tools and how quickly the U.S. cedes ground to Chinese open-source alternatives.

The Open-Source Counterweight

The restrictions are accelerating interest in open-source models, which live in the public domain and can be downloaded, modified and shared by anyone. Unlike private models controlled by a single company, open-weight systems allow for independent auditing and safety testing. A study from MIT Management earlier this year found open-source models accounted for 20 percent of all AI token usage.

Aaron Levie, CEO of cloud content management company Box, said the situation cuts "to the core of one of the central debates" in AI. "If open weights AI can remain a close second to frontier intelligence, then the equation reverses," Levie wrote. "With a highly regulated approach, you'll own the frontier market still, but the vast majority of tokens used will go to an alternative stack."

China's Open-Source Advantage

The regulatory environment may benefit Chinese AI developers, who offer cheaper open-source models to global customers. Research from venture capital firm Andreessen Horowitz shows 80 percent of developers worldwide who use open-source tools are building with Chinese models. Airbnb CEO Brian Chesky said last year the company was "relying a lot" on Alibaba's Qwen model, describing it as "fast and cheap."

Alex Karp, CEO of Palantir Technologies, criticized the token-based pricing model used by frontier labs like OpenAI and Anthropic. "Something has gone completely wrong," Karp said on CNBC's "Squawk Box" on July 1. "What the technical customers want is control over their compute, their models, their data stack and their alpha."

The cost differential is driving enterprise adoption of open models. Frontier model token costs are rising as users maximize usage — a practice known as "tokenmaxxing" — pushing companies toward cheaper alternatives. Vinny Troia, founder of intelligence firm Shadow Nexus, called the pricing structure "price gouging" and said the White House actions "are only going to further compound the problem."

Anthropic CEO Dario Amodei has long opposed open-weight models, warning during a 2023 Capitol Hill visit that scaling open-source AI is "going down a dangerous path." He argued private models allow companies to moderate usage and revoke access when misuse is detected. But the events of the past month have crystallized the debate: when a government can order a model taken down in 90 minutes, the question of who controls access — and under what rules — becomes the central issue for the AI industry.

This article is for informational purposes only and does not constitute investment advice.