OpenAI Joins Anthropic Under Government Access Controls as Fable 5 Remains Blocked

Since the Commerce Department's June 12 export control directive pulled access to both Claude Mythos 5 and Fable 5, two concrete things have changed: Mythos 5 is back for a defined set of U.S. partners, and the same government framework has now reached OpenAI.

The new detail in Friday's developments is OpenAI. The company said the U.S. government asked it to limit preview access for GPT-5.6 to a "small group of trusted partners," according to Business Insider. OpenAI did not describe the request as a formal export control directive, but the pattern is identical to what happened to Anthropic two weeks earlier. The timing is notable: GPT-5.6's restricted release landed the same day Commerce Secretary Howard Lutnick signed the letter partially restoring Mythos 5 access.

OpenAI was direct about its position. "We don't believe this kind of government access process should become the long-term default," the company said in a statement quoted by Business Insider. "It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them."

What the Lutnick Letter Actually Says

Lutnick's June 26 letter to Anthropic's chief compute officer Tom Brown, obtained by The Epoch Times and confirmed by Semafor, authorized Mythos 5 access for entities listed in an attached Annex A and their foreign national employees. Lutnick wrote that Anthropic has "committed to work with the U.S. government on protocols and standards and releases" for its models, and cited "significant progress" from two weeks of daily talks.

Semafor reported the authorized group exceeds 100 U.S. institutions, including major corporations and government agencies. Anthropic described them as organizations that "operate and defend critical infrastructure." The Epoch Times noted that Lutnick's letter did not specify the number of companies or the selection criteria.

Fable 5, the consumer-facing version built on the same underlying model as Mythos 5, remains blocked. Semafor reported that people close to the talks say a Fable 5 release is moving toward approval, but no timeline has been stated publicly. Anthropic's statement said it is "continuing to work with the government" on Fable 5's return to general use.

The Concern That Started This

The strongest case for the government's intervention deserves a fair hearing. Amazon and other companies reportedly warned U.S. authorities that both models could be jailbroken, and Semafor reported that officials were also concerned a China-linked group had gained access to Mythos 5 through a South Korean telecommunications provider. If those concerns are accurate, a frontier cybersecurity model with documented access vulnerabilities reaching adversary-linked entities is a genuine national security problem, not a bureaucratic overreaction.

Anthropic's position, stated at the time of the original June 12 shutdown, was that the government did not provide specific details of its national security concern and that the jailbreak vulnerability was already widely available. That dispute about severity has not been resolved publicly. What has changed is that Anthropic completed two weeks of daily negotiations and made enough progress to satisfy Lutnick's standard for Mythos 5, at least for vetted partners.

A Regulatory Regime Being Built in Real Time

Commerce Department spokesman Benno Kass framed the two-week turnaround as a success: "In just two weeks, we have worked diligently to ensure America remains the global leader in AI while safeguarding our security," according to Semafor. That framing treats speed as validation.

The harder question is what this framework looks like at scale. Semafor noted that the rules are "being built on the fly" and that non-U.S. governments, allied companies, and consumers remain in the dark about when they will get access to Mythos 5 or Fable 5. European officials, according to Semafor, have expressed frustration at their new dependence on Washington's decisions for access to tools their own organizations rely on.

Engadget's coverage noted the original June 12 order barred access to the models by foreign nationals even inside the U.S., including foreign national Anthropic employees, which Anthropic highlighted as operationally unworkable. The Lutnick letter explicitly lifts that restriction for entities on Annex A and their foreign national employees, suggesting that particular concern was resolved in negotiations.

The open question now is whether this vetted-partner model becomes the permanent architecture for frontier AI releases or whether it's treated as a temporary security measure while underlying vulnerabilities are patched. OpenAI's GPT-5.6 is already operating under a similar access restriction one week into its existence, which suggests the former is more likely.