OpenAI previewed GPT-5.6 Sol last Friday, but only to a small group of partners, after the Trump administration asked the company to slow the rollout over safety concerns. Sam Altman said OpenAI complied while warning that this kind of restriction "should not be the norm."
It is the second frontier model in a month to face government gatekeeping after Anthropic's Mythos went through a similar process. The pattern suggests Washington is shifting from passive observer to active broker of which AI tools reach the public.
What can it do?
GPT-5.6 Sol targets coding, science and cybersecurity workloads.
Ships with what OpenAI calls its most advanced safety stack to date.
Available initially to a select group of vetted partners.
Paired with Jalapeño, OpenAI's new inference chip built with Broadcom.
Wider public release timeline left unconfirmed by Altman.
Upgrade to Absolutely Agentic Premium
Join our growing subscriber community and get access to:
Our daily Agentic Intelligence newsletter curates all the best stories from the top AI newsletters, development labs, news websites and social media into one email.
Ad-free podcast - premium ad free RSS feed that you can put into your podcast app of choice. We publish about 3 hours of video a month and now its available in audio form.
Subscriber area includes long form guides to the most important AI trends affecting content and marketing, along with automations and prompts that will save you a lot of time.
Why does it matter?
The model represents a noticeable jump in agentic coding and scientific reasoning, exactly the domains regulators worry about most. Restricting who can use it slows OpenAI's commercial momentum and gives Asian rivals time to close the gap with cheaper, less restricted alternatives.
Government pre-clearance is becoming a soft licensing regime, with both Fable and GPT-5.6 routed through Washington before launch. Smaller labs without OpenAI's political capital may find the same path harder to navigate, concentrating frontier access among a handful of incumbents.
Our take
The Trump administration spent its first year cutting AI red tape and lecturing Europe about overregulation. Yet it’s asking the country's largest AI lab to slow walk a flagship model. Frontier access is now treated as a national security asset, and the White House wants a hand on the dimmer switch even while publicly promoting a light-touch approach. Altman's objection reads as positioning, marking the boundary so OpenAI can argue next time that the precedent has gone far enough.
For enterprises and developers, the practical takeaway is timing uncertainty. A staggered preview means partners with existing relationships, including large clouds, defence contractors and well-connected startups, get a head start of several months on building with the new capabilities. Everyone else waits, and that waiting period is now controlled by an opaque review process rather than OpenAI's product schedule. Capability gaps between insiders and outsiders will widen, and customers will start treating political access as a procurement criterion.
Another big thing… Anthropic's Mythos gets the green light
On Friday the Trump administration loosened restrictions on Anthropic's Mythos 5, clearing it for use by more than 100 US companies and government agencies, including their non-American employees. The decision de-escalates a months-long standoff and gives Anthropic the federal and Fortune 500 distribution it needs to keep pace with OpenAI. Asian labs have already started shipping Mythos-style alternatives without the export friction, so the window for US labs to convert this regulatory thaw into market share is narrowing fast.
