US demands OpenAI restrict GPT 5.6: government control of AI tightens

What happened?

The US government has asked OpenAI to limit the distribution of GPT 5.6, its next frontier AI model, to a small number of partners previously approved by the administration. According to CNN, this request is based on the model's advanced capabilities, which authorities consider a potential risk to national security. The measure is not an isolated event: it comes just days after the Trump administration imposed an export control order on Anthropic, forcing the company to withdraw its Mythos and Fable models from the market. OpenAI has accepted the restriction as a temporary solution to launch GPT 5.6, though it has expressed dissatisfaction with the process.

Why is this important?

This event marks a turning point in the relationship between the US government and leading AI companies. For the first time, direct restrictions are imposed on a frontier model before its public launch, setting a significant regulatory precedent. The Anthropic case shows that Washington is willing to take drastic measures if it considers a model poses a risk to national security or cybersecurity. The decision could accelerate the development of formal regulatory frameworks for frontier AI, affecting the competitive dynamics of the sector. It also introduces uncertainty for enterprise customers, who will now have to undergo a government approval process to access GPT 5.6.

Context and background

The immediate trigger was the episode of Anthropic's Mythos and Fable, models that alerted Washington and Wall Street due to their cybersecurity capabilities. According to sources from Xataka, experts consider Mythos to represent an unprecedented risk vector, with capabilities that could be used for advanced cyberattacks. The administration and OpenAI agree that GPT 5.6 is on par with Mythos in terms of power and scope, making it the second frontier model subjected to direct government scrutiny in a matter of weeks. Sam Altman, CEO of OpenAI, described the situation in an internal report as a "strange moment" without a real regulatory framework, according to The Information. The government is approving access "client by client," a formula OpenAI tolerates but does not endorse, and which could slow the model's adoption.

Consequences for the sector

• OpenAI: The company is forced to accept a controlled launch model that could delay GPT 5.6 adoption and limit its competitive advantage. Altman has expressed concern about the lack of a clear regulatory framework.
• Competitors: Other AI companies, such as Google DeepMind or Meta, could face similar restrictions, affecting competition and innovation in the sector. The measure could concentrate power in a few government-approved players.
• Regulation: The situation could accelerate the development of formal regulatory frameworks for frontier AI, possibly with licensing requirements and mandatory safety assessments before launch.
• Enterprise customers: Companies wishing to use GPT 5.6 will have to go through a government approval process, adding uncertainty and costs. Sectors like defense or finance could benefit from priority access, while others could be excluded.
• Market: The measure could slow the adoption of advanced AI in the US, while international competitors, especially China, could move faster without similar restrictions.

Reactions and next steps

OpenAI has stated that "it is not their preferred model in the long term" and will work with the government to achieve a more sustainable approach, according to internal sources. The White House says it "continues to collaborate with frontier AI labs" to ensure national security. The debate on AI regulation is expected to intensify in the coming months, especially ahead of the 2026 elections. Some experts predict that licensing requirements and mandatory safety assessments for frontier models will be established, similar to export controls on critical technologies. Meanwhile, OpenAI and Anthropic could push for clear rules to avoid the current uncertainty. The case could also inspire other countries, such as the European Union, to strengthen their own AI regulations.