White House Asks OpenAI to Slow Down Release of New AI Model

OpenAI, a leading company in the world of AI, has encountered an unexpected obstacle in presenting its next GPT 5.6 model. The Donald Trump administration is demanding a temporary halt to the public release of this technology and its deployment under strict control for security reasons. This was reported by The Information based on its sources. Techcrunch.com reports as well.

Company CEO Sam Altman informed employees during a meeting this week that the new model would not be distributed openly like previous versions. According to him, GPT 5.6 will initially be presented only to a limited circle of partners. Most importantly, each user's access to this model will be separately reviewed and approved by the government.

State Control and Cybersecurity Measures

Although the Trump administration initially promoted a policy of non-interference in the AI sector, it has decided to strengthen federal oversight in recent months. According to reports, the Office of the National Cyber Director and the Office of Science and Technology Policy are working closely with OpenAI. These agencies aim to test the new model's capabilities for launching cyberattacks.

A recently signed executive order by the US President requires AI companies to voluntarily submit new models for government testing before announcing them to the public. OpenAI is currently operating based on these exact requirements. If the limited testing period is successful, access for the general public is expected in a few weeks.

The Anthropic Experience and Risk Level

The situation OpenAI is facing resembles the actions of its main competitor, Anthropic. Anthropic provided its powerful model, named Claude Mythos, only to selected partners within the Project Glasswing project. At the time, the company argued that the technology "could cause more harm than good if it falls into the wrong hands."

Experts believe that GPT 5.6 and similar advanced models could become powerful tools for cybercriminals. Modern neural networks are capable not only of writing malicious software but also of finding and disabling vulnerabilities in complex systems without human intervention. This poses a serious risk to large corporations and state infrastructures.

For now, the real capabilities of the GPT 5.6 model remain a secret, as it has not yet been studied by independent experts. However, such sharp intervention by the White House may indicate that the capabilities of the new model are significantly higher than previous ones. For Uzbek users and local IT specialists, these restrictions may be a prelude to new regulations in the global AI market.