OpenAI is limiting access to its newest AI model, GPT-5.6 Sol, to roughly 20 pre-approved customers cleared by the Trump administration, a move that underscores how deeply Washington’s scrutiny of frontier AI has begun to shape how companies bring products to market.
Sol, pronounced like the Spanish word for sun, will remain restricted to a narrow set of vetted partners while officials assess its national security implications. OpenAI described the arrangement as temporary, pointing toward broader access in the coming weeks, but was careful to add that it does not want government gatekeeping to become a long-term default.
The episode sits alongside a more disruptive intervention the administration made earlier this month against Anthropic. Two of Anthropic’s new models, Fable 5 and Mythos 5, were pulled offline just days after their debut when a Trump directive blocked foreign nationals from using them. Fable, which Anthropic had positioned as a less dangerous alternative to Mythos, has now been off the market for two weeks.
Sponsored · Altamira Gold Corp.
Much of the administration’s alarm originates with warnings Anthropic itself issued about Mythos. David Sacks, who co-leads Trump’s council of technology and science advisers, recounted on a recent podcast that Anthropic CEO Dario Amodei traveled to Washington in April and described Mythos as effectively a cyber weapon capable of finding software vulnerabilities at scale.
“He spiked the cortisol level, got everyone really worried,” Sacks said.
Trump signed an executive order in June establishing a framework under which the federal government can subject the most powerful AI models to national security review for a period of up to 30 days before those products reach the public, with developer participation described as voluntary.
OpenAI has navigated the administration’s concerns more smoothly than its rival. CEO Sam Altman spoke directly with Commerce Secretary Howard Lutnick about the Sol release last week, part of a broader series of negotiations between tech executives and Trump officials. Anthropic’s path has been far rougher. The Pentagon designated the company a national security risk after it raised ethical objections to AI use in warfare, Trump ordered federal agencies to drop Claude, and the company filed a lawsuit against the administration that remains unresolved in federal court.
On the technical substance, OpenAI acknowledged Sol carries some cybersecurity risk, particularly when combined with other tools, but argued it does not cross the company’s internal threshold and is more useful for finding and patching vulnerabilities than for exploiting them.
Information for this story was found via the sources and companies mentioned. The author has no securities or affiliations related to the organizations discussed. Not a recommendation to buy or sell. Always do additional research and consult a professional before purchasing a security. The author holds no licenses.
