White House Moves to Restrict OpenAI's GPT 5.6 Release Amid Broader AI Oversight Push

Government Seeks Limited Release for Advanced AI

The White House has formally requested that OpenAI restrict the public release of its forthcoming GPT 5.6 model. This directive, reported on June 25, 2026, highlights growing governmental concern over the advanced capabilities of leading artificial intelligence systems. The Trump administration is asking OpenAI to limit access to GPT 5.6 to a select group of government-approved partners, citing safety considerations and the model's powerful nature. This specific request to OpenAI follows a broader executive order signed earlier in June, which aims to bolster national security and cybersecurity in the rapidly evolving landscape of AI technology.

Executive Order Mandates Early Access and Vetting

President Donald Trump's executive order, signed on June 2, 2026, establishes a framework for potential government oversight of new AI models before their public launch. The order encourages AI developers to voluntarily participate in benchmarking processes designed to assess a model's "advanced cyber capabilities" and determine if it qualifies as a "covered frontier model." For models classified as such, the executive order requests access for up to 30 days prior to their general availability. This pre-release access would enable the government to select "trusted partners" who could then utilize these models to foster secure innovation and enhance the cybersecurity of critical infrastructure.

While the administration has clarified that this order is not intended to create a mandatory governmental licensing or preclearance system for AI model development or distribution, it also directs the Department of Justice to prioritize enforcement of laws against the use of AI for cyberattacks or other criminal activities. This includes unauthorized access to information technology systems or data. This proactive stance was anticipated by reports in May 2026, which indicated discussions within the Trump administration about forming a working group of industry executives and government officials to develop options for federal government vetting of new AI models. Earlier that month, major AI developers including Google DeepMind, Microsoft, and xAI had already agreed to share their frontier AI models with a government agency for national security testing.

Shifting Policy and Industry Dynamics

This federal intervention marks a notable shift for the Trump administration, which had previously been perceived as dismantling AI safety frameworks established by the preceding Biden administration. The government's increasing demand for early access reflects growing public and official concern about the potential risks posed by advanced AI. Industry observers note that the current AI safety regime largely depends on the discretion of a few executives at leading AI companies, as exemplified by Anthropic's decision to withhold its Mythos model due to cybersecurity risks.

The concentration of AI research capacity within the private sector further complicates independent oversight. Approximately 80 percent of global AI computing power is privately owned, and nearly 70 percent of new AI PhD graduates are recruited directly into the private sector, a significant change from a decade ago when universities led much of the AI research. This imbalance underscores the challenges for independent evaluation and the government's push for early insights into cutting-edge AI developments.

The administration's dual approach of an executive order and specific requests to developers like OpenAI signals a growing recognition of AI's dual-use potential, prompting a necessary re-evaluation of how innovation and security are balanced in this rapidly advancing field.