OpenAI, the company that created ChatGPT, announced on Friday (June 26, 2026) the postponement of the public launch of its newest artificial intelligence model, GPT-5.6. The decision came after a formal request from the United States government, which seeks early access to assess potential national security risks.
The measure restricts initial access to the AI model to a select group of verified partners, whose data has been shared with authorities. This move underscores growing concern in Washington about the dangers posed by powerful artificial intelligence systems.
National security concerns drive American measure
The US government’s request reflects growing apprehension in Washington about the dangers that advanced AI systems could pose. Lawmakers and authorities have intensified pressure on developers to implement strict safeguards before making their technologies available to the public.
The main purpose of this early access is to identify potential threats, from vulnerabilities to complex cyberattacks to inappropriate use by military or malicious actors. The intention is to mitigate these risks before artificial intelligence tools become widely accessible.
OpenAI collaborates on temporary framework for new AI releases
In an official post, OpenAI clarified that the restricted release of GPT-5.6 is an interim step. The company is collaborating with Washington to develop a more comprehensive regulatory framework for future launches. The model’s plans and capabilities were previously presented to the government.
This short-term measure, according to the company, aims to pave a more solid path for the model’s wide availability in the coming weeks. The company seeks, together with the government, to establish a replicable structure for the next models, following cybersecurity guidelines.
Donald Trump’s executive order already provided for early access to cutting-edge models
The current scenario is in line with an executive order signed by President Donald Trump at the beginning of the month. This executive order establishes a voluntary framework that allows artificial intelligence developers to offer “covered frontier models” to the United States government for up to 30 days. This period precedes release to trusted partners.
The initiative demonstrates a proactive strategy by the government to ensure that emerging technologies are properly evaluated before their widespread implementation. A delicate balance is sought between technological innovation and the strategic vigilance necessary to protect national interests.
OpenAI Challenges and Concerns with Access Restriction
While OpenAI reinforces its commitment to rigorous testing and coordination with partners, the company has expressed concern about the possibility that this level of government oversight will become a permanent standard. The company did not reveal the names of the partners who already have access.
There is expressed concern that overly restrictive regulatory processes could limit access to advanced artificial intelligence tools. This would affect crucial users such as developers, enterprises, cybersecurity professionals and international partners, who could significantly benefit from these innovations.
Details of OpenAI’s new AI models
OpenAI’s next generation of artificial intelligence includes three distinct models, each with a specific focus:
- GPT-5.6 Sol:It represents the most advanced model developed by the company to date, promising unprecedented capabilities in various applications.
- Earth:An intermediate-level model, seeking to balance performance and cost-benefit for a wider audience of developers and companies.
- Luna:Designed to be the most affordable option, aiming to democratize access to advanced artificial intelligence technologies.
Regulatory precedent: the case of Anthropic and government control
This is not an isolated incident of government intervention in the artificial intelligence sector. Earlier this month, the United States government had already issued an order to OpenAI competitor Anthropic, demanding the suspension of foreign citizens’ access to its cutting-edge AI models.
The justification for the action against Anthropic also involved direct national security concerns. Currently, the creator of Claude remains involved in a legal and regulatory battle with American authorities, indicating a growing trend of scrutiny over the development and distribution of AI technologies.
Amid these regulatory developments, both OpenAI and Anthropic have confidentially filed initial public offering (IPO) applications in the United States. There are reports that OpenAI is considering postponing its stock market debut until next year, in a scenario of intense government and market evaluation.