White House asks OpenAI to delay GPT-5.6 release over safety fears

The Trump administration has directed OpenAI to limit the initial release of GPT-5.6 to government-approved customers, a first-of-its-kind intervention that signals a hardening of federal oversight over frontier artificial intelligence models.

"We've made clear to the US government that this is not our preferred long term model, and will work with them and others in industry to achieve a more sustainable approach for future releases," Sam Altman, chief executive officer of OpenAI, told employees in a company Q&A, according to The Information.

Under the arrangement, the administration will approve customer access on a case-by-case basis during a preview period expected to last several weeks before a broader public release. Several agencies are involved, including the Office of the National Cyber Director, the Office of Science and Technology Policy, and Commerce Secretary Howard Lutnick. The directive follows an executive order President Donald Trump signed earlier this month asking AI companies to participate in a voluntary federal review of powerful models before public release.

The intervention raises the stakes for the entire AI industry. OpenAI rival Anthropic earlier this month received a far stricter ultimatum requiring it to suspend access to its Mythos 5 and Fable 5 models and bar foreign nationals — including its own non-US employees — from using the technology. The uneven treatment between the two leading labs has created confusion about how the review process works and how voluntary it actually is.

The GPT-5.6 delay comes as the battle to govern artificial intelligence intensifies across multiple fronts. At the G7 Summit in Evian, France, world leaders elevated AI alongside trade and national security as a defining strategic issue, with frontier AI companies participating directly in diplomatic discussions for the first time. Pope Leo XIV has described AI as one of the defining moral challenges of the current generation, arguing that technological progress must never come at the expense of human dignity.

Five centers of power are emerging

The governance environment is fragmenting into five distinct centers of influence, each with different priorities. Political authority rests with governments seeking to protect democratic institutions. Technological authority resides in Silicon Valley labs such as OpenAI, Google DeepMind, Anthropic and Meta, whose decisions increasingly shape national security and global economics. Moral authority has been asserted by the Vatican, while national strategy is driven by the White House's focus on economic competitiveness and military superiority. Meanwhile, strategic competitors including China, Russia and North Korea are advancing AI capabilities with little regard for Western governance debates.

Cybersecurity as the collision point

These forces converge most sharply in cybersecurity. Agentic AI systems capable of planning and executing complex tasks with limited human intervention are exposing new vulnerabilities. Meta this week absorbed the founding team of AI safety startup Virtue AI — co-founders Bo Li, Dawn Song and Sanmi Koyejo — into its Fundamental AI Research Lab to fortify its autonomous agents against emerging threats. "As we ship AI products to billions of people and build increasingly capable agents, keeping those systems safe, reliable, and trustworthy is foundational," Meta said in an internal communication obtained by Axios.

The OpenAI directive raises unresolved questions about accountability and authority in the AI era. How much autonomy should AI systems possess? Who remains accountable when autonomous systems make the wrong decision? And how do democratic societies preserve accountability when adversaries operate without similar constraints? These questions will be answered over the coming decade through collaboration, competition and tension among governments, technology companies, moral institutions and security leaders.

This article is for informational purposes only and does not constitute investment advice.