Published on 6/27/2026
The US government is preparing to impose massive control over the artificial intelligence models that are released, after it decided to restrict the availability of the new “GPT 5.6” model of the company OpenAI to most users, and before that, restricting the models of “Fable” and “Mythos” affiliated with Anthropic.
The American newspaper “The Information” explained that the White House informed OpenAI of its desire for the company to launch its next model in a limited manner and to limit it to a selected group of close partners.
The company’s CEO, Sam Altman, reportedly told employees that their latest GPT 5.6 model would be rolled out in a completely different way than previous models, with the government agreeing to grant access “one customer at a time.”
According to the newspaper’s report, after this limited release period, the company should be able to launch the model on a larger scale after “two weeks,” according to expectations.
GPT 5.6 represents a significant improvement over GPT 5.5, both in terms of context window size and in terms of efficiency. In a memo sent to employees, Altman explained that OpenAI will work with the government and other parties in the technology sector to come up with a more sustainable approach to future releases.
Fable 5 and Mythos 5
This action comes after OpenAI’s competitor Anthropic was recently forced to withdraw its more powerful model, “Fable 5”, after the administration of US President Donald Trump intervened to prevent the model from reaching foreign hands. The company had previously launched the “Mythos 5” model – a more powerful model – as a limited edition available only to a small group of pre-approved customers.
For its part, the American technical website “The Verge” indicates that if this preview lasts only “a few weeks”, this may not represent a major problem, but the Mythos model has been going through the preview stage for months without any indication that a public version is imminent.
Even if the review takes only a few weeks, it could significantly limit the economic returns of expensive new systems at a time when AI companies are trying to improve their bottom lines. If the pace of model development slows as a result, it will likely freeze ongoing data center construction. If things get worse, the entire industry could be in real danger.
A common fate
OpenAI and Anthropic now find themselves in exactly the same situation, facing the same problems and the same potential disaster if they fail, according to The Verge. Discussions within the technology sector usually focus on the role of one side in bringing about this crisis, whether by accusing Anthropic of running a regulatory takeover scheme, or accusing OpenAI of cozying up to Trump to alienate its competitor.
According to the same site, the most pressing problem is finding a logical issuance process. Although it is acceptable for the government to test models before launching them, which is what it does with many consumer products, it is not clear what kind of safety guarantees can be put in place to satisfy regulators.
The Verge adds that the US government does not have the experience or ability to conduct this type of necessary testing, and it is not clear what regulators are trying to protect against, as there have been no efforts to clarify the risks that the government is actually concerned about.
Mythos 5
The new decision to restrict the OpenAI model comes at the same time as it decided to grant Anthropic permission to republish the “Methos 5” model for “a group of American organizations that manage and protect critical infrastructure,” according to what the company announced on the X platform.
Anthropic announced yesterday, Friday, that it had obtained a permit from the US government allowing a small group of cybersecurity companies in the United States to access its advanced model of artificial intelligence, “Methos 5.”
According to a report by the Semafor website, the company obtained permission to redeploy its most powerful cybersecurity models to more than 100 organizations in the United States, including major companies and government agencies.
An Anthropic spokesman said that the company, whose relationship with the Trump administration has been turbulent for months, would continue talks with the government to “expand access to the Mythos 5 models and re-make the Fable 5 model available to the public.”
In this context, “The Verge” website believes that restricting the issuance of artificial intelligence models will not be the complete solution because it will only limit what is available to the public, and that despite the legitimate concerns of the American government about the revolution brought about by artificial intelligence tools in the field of cybersecurity, the best way to address these concerns is to work together and line up behind the least bad regulatory options.
