The White House has asked OpenAI to slow the launch of GPT‑5.6, limiting early access to vetted partners while the government evaluates safety risks under a new executive order.
OpenAI’s newest model, GPT‑5.6, will not hit the public market until the White House clears it. The company has agreed to a limited preview, giving only a select group of partners access while the government conducts a safety review.
The decision follows a June executive order that directs AI firms to submit new models to federal agencies for testing before public release. OpenAI’s CEO, Sam Altman, told staff that the government would approve access on a customer‑by‑customer basis during the preview period. If the limited rollout proves safe, the company plans a broader release a couple of weeks later.
The agencies involved in the review are the Office of the National Cyber Director and the Office of Science and Technology Policy. According to a report by TechCrunch, the administration’s request mirrors Anthropic’s voluntary “Project Glasswing,” which keeps its most powerful model, Claude Mythos, behind a partner‑only gate.
The Trump administration’s push for oversight began with an executive order in December 2025 that created a Department of Justice AI Litigation Task Force to challenge state AI laws. The order also instructed the Federal Trade Commission to issue a policy statement on AI systems that distort outputs for ideological purposes. The policy, open for comment until July 31, would affect how the FTC regulates AI companies.
OpenAI’s compliance signals a shift in the industry’s approach to safety. The company’s spokesperson said the limited release would allow the government to evaluate potential misuse and bias before a wider rollout. The move also reflects growing pressure from lawmakers and regulators to curb the rapid deployment of powerful models.
Industry observers note that Anthropic’s earlier decision to restrict Claude Mythos was partly a marketing tactic, but it also highlighted the real risks of untested AI. The new federal framework could force other firms to adopt similar gatekeeping practices, slowing the pace of innovation but potentially reducing harm.
The policy’s impact extends beyond OpenAI. Meta’s recent decision to charge developers for its most advanced model, announced by DIGITIMES, is part of a broader trend of monetizing frontier AI to fund safety research. Meanwhile, Google’s Gemini team, as reported by Android Authority, is actively soliciting user feedback on reliability and tool calling, indicating that companies are already grappling with the same safety and usability challenges.
For developers, the new rules mean a longer lead time before they can access cutting‑edge models. For consumers, the delay could reduce the risk of encountering biased or harmful outputs in everyday applications. The balance between speed and safety remains a contentious point in the AI debate.
The White House’s insistence on a safety review raises questions about the future of AI regulation. Will federal oversight become the norm for all large‑scale models, or will it be limited to a handful of high‑profile releases? The answer will shape the competitive landscape for years to come.
FAQ
*What is GPT‑5.6 and why is it significant?* GPT‑5.6 is OpenAI’s next large‑language model, expected to surpass GPT‑4 in size and capability. Its release is closely watched for potential applications in medicine, finance, and creative industries.
*How does the government’s review process work?* The government will evaluate the model’s safety, bias, and potential misuse through a series of tests and audits before granting broader access.
*Will other AI companies be required to follow the same path?* The executive order applies to all AI firms, but the extent of enforcement and compliance deadlines are still being clarified.
*What are the risks of delaying the release?* Delays could slow innovation and give competitors a relative advantage, but they may also prevent the deployment of models that could cause harm if misused.
