Tech

White House asks OpenAI to slow down its next AI launch, signalling a new phase in Washington’s safety concerns

White House asks OpenAI to slow down its next AI launch, signalling a new phase in Washington’s safety concerns

“For years, the debate was whether governments could keep up with AI. Now the US government is actively asking one of the industry’s biggest players to hold back a major release before the public ever gets access.”

OpenAI is reportedly being asked by the Trump administration to slow the rollout of its upcoming AI model, a request that highlights growing concerns inside Washington about how powerful new artificial intelligence systems should be tested and controlled before reaching the public.

The model at the center of the discussion is GPT-5.6, which OpenAI had been preparing to release more broadly. Instead, the company is now expected to begin with a limited rollout to a small group of approved partners rather than making the model widely available from day one. The move follows direct requests from officials concerned about the potential risks associated with increasingly capable AI systems.  

According to reports, the White House, along with several federal agencies, wants additional time to evaluate the model’s capabilities and security implications before a wider launch takes place. The administration’s concerns reportedly focus on national security, cybersecurity risks, and the possibility that advanced AI systems could be misused if released too quickly.  

The request marks a notable moment for the AI industry.

For much of the past year, Washington has struggled to balance two competing goals: maintaining American leadership in artificial intelligence while also addressing concerns about safety and misuse. The administration has generally promoted rapid AI innovation, but recent developments suggest officials are becoming more cautious as models grow more powerful.  

OpenAI appears willing to cooperate.

SEE ALSO: OpenAI unveils its first custom AI chip as it tries to loosen Nvidia’s grip

Reports indicate that CEO Sam Altman and senior company executives have been working with government officials on a framework that would allow limited early access while additional testing takes place. Under the reported arrangement, only a select number of organizations would receive the model initially, with broader availability potentially coming weeks later if officials are satisfied with the results.  

The development also comes against the backdrop of growing government scrutiny of frontier AI models.

Earlier this month, rival AI company Anthropic faced restrictions affecting access to some of its most advanced systems after concerns were raised about its capabilities. Those actions sparked debate throughout the technology industry about whether governments should have the authority to limit access to cutting-edge AI tools and how such decisions should be made.  

That context makes the OpenAI situation particularly significant.

Unlike traditional software releases, advanced AI models are increasingly being treated by policymakers as strategic technologies with national security implications. Officials are worried not only about misinformation and cyber threats, but also about how rival nations could potentially exploit powerful AI systems if safeguards are insufficient.  

Critics, however, argue that government involvement in model releases could create uncertainty for the industry and raise questions about who gets access to powerful technology and who decides when it is safe enough for public use. Some observers have also pointed to apparent inconsistencies in how different AI companies are being treated by regulators.  

For now, OpenAI’s next major model appears headed for a more cautious debut than many expected.

The bigger story is not just about GPT-5.6 itself. It is about a changing relationship between Silicon Valley and Washington, where the release of a new AI model is increasingly being viewed less like a product launch and more like a matter of national policy.  

Filed under: Tech