Mythos 5, the strongest AI on the planet, is finally released! According to reports from many authoritative media including Reuters and Semafor, the U.S. Department of Commerce officially sent a letter to Anthropic on Friday (June 26), approving it to release the most powerful model Claude Mythos 5 to more than 100 "trusted partners".


Unfortunately, although Mythos 5 has been released, ordinary people still cannot use it.

The unblocked list covers many Fortune 500 companies and government agencies, but ruthlessly excludes the vast majority of users around the world, including traditional allies of the United States.


And on the same day, OpenAI also released GPT-5.6 to the government-approved list with great tacit understanding.

100 “Chosen Sons”:

Revealing the Mysterious Annex A Whitelist

Since Claude 5 was completely banned, after 14 days of intense gaming and backroom negotiations, the two sides finally reached a compromise today.

U.S. Commerce Secretary Howard Lutnick wrote in a letter to Anthropic Chief Computing Officer Tom Brown that the efforts of both parties have made "significant progress" and Anthropic has committed to working closely with the U.S. government on the release of protocols, standards and models.


According to Lutnick's letter, the unbanning of Mythos 5 comes with an extremely stringent condition: the export license exemption only applies to entities listed in Appendix A of the letter.

In other words, the United States has officially established a new AI regulatory system - control over the release of cutting-edge AI models has been completely transferred from the hands of Silicon Valley geeks to the desks in Washington.

Under the new rules, Mythos 5 may be exported, re-exported, or transferred domestically to entities listed in Appendix A and their foreign employees, as well as Anthropic’s foreign employees.

This mysterious white list includes more than 100 domestic institutions, many Fortune 500 companies, and key government agencies.


Foreign employees of these whitelisted companies, as well as Anthropic’s own foreign employees, will be allowed access to the model (exempt from “deemed export” restrictions).

For companies that fail to make the list, even if you are in Silicon Valley, even if you are a major paying customer of Anthropic, the ban is still in effect.

Currently, there is no specific and complete public list of whitelists.

It is reported that the whitelist is expanded based on Anthropic’s existing Project Glasswing partner framework.

Currently known core partners of early Project Glasswing include: AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, Linux Foundation, Microsoft, NVIDIA, Palo Alto Networks and other technology and security giants.

Department of Commerce spokesman Benno Kass seemed quite proud of this: "In just two weeks, we have worked around the clock to not only ensure the United States' global leadership in the field of AI, but also to defend our national security."

Pending Fable 5

In this letter of lifting the ban that caused shock in the technology circle, one name was deliberately ignored-Fable 5.

As the "cousin" of Mythos, although the parameter scale of Fable 5 is slightly weaker, before it was banned, it was the most powerful AI model accessible to ordinary consumers in the world.

Its removal from the shelves has made countless developers and C-side users sad.

Currently, the Commerce Department’s letter makes no mention of Fable 5.

Although people close to the negotiations revealed that the government is working towards unbanning Fable, the specific timetable remains a mystery.

This strategy of "protecting B-side giants and government agencies first, and then managing the C-side masses" once again proves that in the face of security strategies, the needs of ordinary consumers can only be postponed indefinitely.

OpenAI’s “amazing coincidence”: GPT-5.6 is also coming

Coincidentally, at the same time today, Anthropic’s biggest enemy, OpenAI, also took action.

They released the latest generation, GPT-5.6, to a small group of government-approved trusted partners!

Technology analyst Andrew Curran exclaimed on social media, expressing the concerns of many industry insiders:

This is so ominous.

If the widespread release of GPT-5.6 ends up becoming US-only, it will be an event of seismic proportions. This also almost guarantees that Fable 5 will face the same fate.

I expected this day to come sooner or later, but definitely not in the summer of 2026. Not so fast. Everything is accelerating.


If everyone was just guessing before, then the two boots that landed at the same time today have completely confirmed the disturbing rumors: the United States is completely privatizing and weaponizing the top cutting-edge AI models.

Under this operation, god-level AIs like Mythos 5 and GPT-5.6 are no longer simple commercial products, but the most lethal diplomatic weapons in the hands of the United States.

The global AI ecosystem is divided

Today, European officials have expressed strong dissatisfaction. They cannot accept that European companies and scientific research institutions will depend on Washington's face before they can use the most cutting-edge AI tools.

This dissatisfaction and anxiety is giving rise to an irreversible trend: the global AI ecological division.

On the one hand, non-U.S. regions may accelerate their shift to China’s AI models, or invest heavily in building their own technology stacks.

On the other hand, the global business models of OpenAI and Anthropic are suffering heavy losses.

In the past, they charged global users membership fees of $20, $50 or more. In the future, this kind of subscription system that treats everyone equally will no longer exist.

Instead, there are tiered services that are customized for different countries and customers and full of additional conditions.

The release of Claude Mythos 5 is a carnival belonging to individual giants and institutions. But for hundreds of millions of developers, researchers and ordinary users around the world, all they feel is a deep sense of powerlessness.

June 26, 2026, this day is destined to be recorded in history.

It marks the first stage of AI development - the "classical AI era" that was dominated by the spirit of the Internet, open and shared, and enjoyed by global developers, has officially come to an end.