Two of the biggest AI companies are now limiting access to their newest models at the Trump administration’s request. OpenAI said Friday that GPT-5.6 Sol will only be available to government-approved customers. Anthropic got partial clearance to redeploy its Mythos 5 model to a small group of cyber defenders and infrastructure providers, but its other model, Fable 5, remains restricted.
This all stems from a June executive order that created a framework for the federal government to vet advanced AI systems for national security risks — up to 30 days before public release. Participation is technically voluntary, but the framework has not been fully developed yet, and the practical effect has been anything but optional.
The trouble started when Anthropic warned earlier this year that its Mythos model was particularly good at finding software vulnerabilities — good enough that it could be weaponized by malicious hackers. That warning, delivered to Washington in April, spooked officials. Trump adviser David Sacks said on a podcast that Anthropic CEO Dario Amodei “spiked the cortisol level” of policymakers by describing Mythos as a kind of cyber weapon.
OpenAI’s Sol model, by contrast, is positioned as better at helping people find and fix bugs than at launching attacks. The company says it does not cross its own risk threshold. Still, OpenAI acknowledged that combining Sol with other tools could create unforeseen risks, which is why it is doing a phased release with roughly 20 approved customers so far.
Critics are pushing back hard. Stanford cybersecurity expert Alex Stamos reviewed Amazon’s analysis of Fable and said he found no risks that are not already present in publicly available models — including Chinese ones. “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.
Representative Lori Trahan, a Massachusetts Democrat, called the approach arbitrary: “No law. No process. No oversight. Just appointees in Washington deciding who is in and who is out.”
Both companies are exploring IPOs, and this government scrutiny adds another wrinkle. Trump has floated the idea of the US government owning a stake in leading AI firms. For now, the phased releases continue, with both OpenAI and Anthropic saying they are working toward broader availability.
