OpenAI Has New AI Models. Here’s Why You Can’t Use Them
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI is delaying the public release of GPT-5.6 at the Trump White House's request, sharing models first with a small set of US-government-preapproved customers
- The company said it hopes to make GPT-5.6 available to everyone in coming weeks, called the process a short-term step that "keeps the best tools from users," and plans to broaden customers next week to include some international partners
- Trump's executive order earlier this month created a "voluntary process" for AI labs to share models with the government 30 days ahead of broader release, with a carve-out against becoming a de facto licensing regime
- OpenAI executives said no voluntary framework yet exists, putting frontier AI labs in a "weird interim period" where cooperating with the government on launches doesn't feel voluntary
- Two weeks earlier, the White House sent an export control directive to Anthropic, prompting it to take its most advanced models offline for all customers — a standoff WIRED reports is still unresolved
- GPT-5.6 ships in three variants: Sol (most capable on cybersecurity, biology, and agentic benchmarks), Terra (middle-tier), and Luna (fast and affordable)
Why it matters: The delay turns a supposedly "voluntary" cybersecurity review into a gatekeeping step before any major US AI model launch. With OpenAI explicitly noting that no framework exists, frontier labs are now negotiating rollout conditions case-by-case with the White House — a dynamic mirrored by the still-unresolved White House standoff with Anthropic over export controls on its most advanced models.
Ask SkimNews
