OpenAI rolls out Sol amid opaque US government safety reviews

OpenAI is rolling out Sol, its latest advanced LLM, for wide public access. Sol is considered to be at least on par with Anthropic’s Fable, but the exact process the U.S. government used to approve either model for release remains unclear. Mina Narayanan, a senior research analyst at Georgetown’s Center for Security and Emerging Technology (a Washington-based security and technology think tank), told TechCrunch: “Frankly, I don’t have visibility into those exact processes, so yes, I don’t feel like I have enough information to say whether they’re adequate or not.”

Eighteen months into the Trump administration, there is still no clear framework for licensing frontier models. An executive order published last month laid out a roadmap for evaluating them, but the specifics have yet to be filled in; the order instructs six cabinet agencies to determine a final process by early August. The Department of Commerce’s Center for AI Standards and Innovation, the U.S. agency now taking the lead on AI standards, appears to be handling reviews in the interim. Sriram Krishnan, who advised the White House on AI until last month, has said there will not be a centralized AI regulator modeled on the FDA, the U.S. food and drug agency. Dean W. Ball, a former Trump policy advisor now at OpenAI, wrote in his newsletter last month that nobody knows what the requirements are to get licensed.

Critics argue the ad hoc, secretive process leans on personal connections rather than independent safety researchers. Andy Konwinski, a computer scientist who co-founded Databricks, Perplexity, and the Laude Institute, calls the uncertainty an existential problem, arguing that beyond safety, the crux is who holds the power to decide who gets access. That concern is sharpened by the backdrop: Altman reportedly offering as much as 5% to OpenAI’s equity for the administration’s so-called “Trump Accounts,” and Brockman’s role as the largest publicly known donor to Trump’s mid-term political operation. Anthropic’s Fable, meanwhile, was briefly pulled from wider access when the U.S. government forbade its use by foreign nationals, partly because of real concerns about users jail-breaking the model to access hacking capabilities. University of Wisconsin-Madison professor Remzi Arpaci-Dusseau said last week at the Open Frontier conference that there is little sense that responsible people are driving these changes forward.

In a late June blog post, OpenAI said it does not believe this kind of government access process should become the long-term default, adding it would work with regulators on a different path. Konwinski favors an open-commons model, pointing to institutions like the FDA, the NIH, and the national laboratories that bring researchers, officials, and companies together to reach safety consensus; Ball similarly argues progress depends on third-party auditing organizations, licensed by the government, to evaluate frontier labs’ safety practices. Some of that tension traces to industry incentives, playing out in court during Elon Musk’s lawsuit challenging OpenAI’s corporate structure. At the same conference, Two Sigma founder David Siegel warned against a future in which a small number of firms control the technology, the government evaluates it inside secretive laboratories, and the public and scientific community have no access to any of it.

Why it matters

The opaque rollout of frontier models like Sol and Fable highlights a lack of standardized regulatory processes in the U.S., raising concerns about backroom political influence and the exclusion of independent safety researchers from decisions that shape public access to powerful AI.