Nobody Knows How the US Greenlights Frontier AI Models
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI's Sol is rolling out for wide public access, considered at least on par with Anthropic's Fable; CEO Sam Altman told CNBC the release process involved conversations with Commerce Secretary Howard Lutnick, Treasury Secretary Scott Bessent, and national cyber director Sean Cairncross, but OpenAI declined to share details on who actually tested the models or how.
- Anthropic's Fable was briefly pulled from public access when the US government forbade its use by foreign nationals, partly over jailbreak-to-hacking concerns and partly due to personality clashes with the Trump administration — an episode that may have made OpenAI more cooperative with the government's unknown requests.
- Three outside experts — Georgetown's Mina Narayanan, former Trump policy advisor Dean W. Ball (now working for OpenAI), and Databricks/Perplexity co-founder Andy Konwinski — told TechCrunch they have no visibility into the licensing requirements; Konwinski called it "existentially a problem" of who gatekeeps decisions.
- A Trump executive order laid out a roadmap for evaluating frontier models but left specifics blank, with former White House AI advisor Sriram Krishnan telling the FT "there will not be an FDA for AI"; six cabinet agencies are supposed to finalize a process by early August, with Commerce's Center for AI Standards and Innovation currently taking the lead.
- The political backdrop includes reports that Sam Altman offered as much as 5% of OpenAI equity to the administration's "Trump Accounts" and that OpenAI president Greg Brockman is the largest publicly-known donor to Trump's midterm political operation — making it hard, the article argues, to separate those ties from the government's lighter-touch approach to Sol.
- Konwinski and Ball both pointed to third-party auditors, "open commons" research bodies, or models like the NIH and national labs as ways to bring disinterested safety, alignment, and interpretability researchers into a process they say is now dominated by industry insiders.
Why it matters: Until at least early August — when six cabinet agencies are supposed to finalize a frontier-AI review process — the most powerful AI systems are being released with no defined public scrutiny, and the people shaping the rules include a former Trump advisor now working for OpenAI and an industry that has donated heavily to the administration. Sol's release sets the template: private conversations, undisclosed testers, and no independent oversight.
