U.S. Approves OpenAI, Anthropic Models One by One
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- The U.S. government pulled Anthropic's Fable and Mythos models two weeks ago, and now OpenAI's GPT 5.6 faces the same limbo with the government set to approve its limited preview "customer by customer."
- The Information broke the GPT 5.6 story Thursday, reporting that a general release depends on per-customer government sign-off, with Altman reportedly projecting a "couple of weeks" of review.
- Anthropic's Mythos has already been in preview for months without a general release timeline, showing how even short regulatory delays can stall costly new systems.
- GMU fellow Dean Ball, described as a soon-to-be OpenAI employee, argued that the U.S. government lacks the expertise and capacity to test frontier models and hasn't articulated what specific risks it's guarding against.
- OpenAI and Anthropic now face identical regulatory exposure, with the source arguing any fix to the approval process inevitably benefits one lab as much as the other — reframing the moment as an industry-wide crisis rather than a rivalry.
- Beyond the process itself, the source flags real underlying concerns about AI's impact on cybersecurity, biorisk, and alignment that restricting releases alone cannot solve.
Why it matters: Both OpenAI and Anthropic now face the same U.S. government approval bottleneck for frontier model releases, and the source argues any solution benefiting one lab inevitably benefits the other. With Anthropic's Mythos stuck in months-long review and GPT 5.6 facing customer-by-customer sign-off, the source notes that even short delays could significantly limit the economic upside of costly new systems and chill the ongoing data center buildout.
