US Allows Anthropic to Restore Mythos 5 Access for 100 Trusted Partners - Fable 5 General Access Still Blocked
The two-week standoff between Anthropic and the US government reached a partial resolution on June 26, 2026. Commerce Secretary Howard Lutnick sent a letter to Anthropic confirming that approximately 100 companies and federal agencies will be permitted to access Claude Mythos 5 - the company's most powerful cybersecurity AI model - under new safeguards. Fable 5, the publicly available model that was disabled alongside Mythos 5 on June 12, remains unavailable while discussions continue.
The US government on Friday granted Anthropic permission to release its Mythos 5 model to a group of roughly 100 companies and federal agencies. "I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model," US Commerce Secretary Howard Lutnick wrote in the letter. "Since the issuance of my June 12 letter, Anthropic has worked with the US government to address risks associated with the Covered Models. These efforts have yielded significant progress."
Anthropic said: "We received notice from the US government that Mythos 5, our strongest cybersecurity model, can be redeployed to a small group of cyber defenders and infrastructure providers. We are working to provision the approved set of providers and restore their access to Mythos 5 as quickly as possible. We are pleased to see this progress and continue to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again."
What the Partial Resolution Establishes
The Lutnick letter is notable for what it preserves as much as what it restores. The letter notably does not mention any change to the restrictions for the Fable 5 model. Lutnick also wrote: "I reserve the right to reevaluate and adjust the scope of license requirements on the Covered Models, should circumstances change."
That reservation clause means the approved list of 100 trusted partners can be expanded or contracted at the Commerce Department's discretion. Anthropic has not regained autonomous control of its own model deployment. The government has established itself as the gatekeeper for who can access the most capable Anthropic models - a precedent with no clear precedent in civilian technology history.
The partial restoration suggests Washington may be settling on a familiar answer from export-control history: keep the most sensitive capability available, but only inside a vetted circle. When a model can be launched on June 9, frozen on June 12, and partially reopened on June 26 under what appears to be a government-approved whitelist, buyers are no longer purchasing only performance. They are buying geopolitical risk, regulatory durability and domestic access.
The OpenAI Parallel
The Anthropic resolution arrived the same day OpenAI announced three new models - GPT-5.6 Sol, Terra, and Luna. Significantly, OpenAI voluntarily followed a government-requested limited rollout for its most capable new model.
OpenAI said in a blog post that it "believes in broad access" and is working to make the models generally available in the coming weeks. OpenAI said it previewed the models' capabilities and shared its plans with the government ahead of Friday's launch. Of the two companies, Anthropic has had a harder time dealing with the Trump administration.
The contrast between Anthropic's experience - forced shutdown, weeks of negotiations, partial restoration under government supervision - and OpenAI's experience - voluntary pre-launch government consultation, managed rollout, no forced shutdown - illustrates the practical value of the political capital that OpenAI has built in Washington. OpenAI hired Dean Ball, the co-author of the Trump AI Action Plan, just weeks ago specifically to strengthen that relationship. The Anthropic situation validated exactly why that hire was necessary.
What Remains Unresolved
Fable 5, the publicly available version of Mythos built with cybersecurity capabilities removed, remains globally disabled. Hundreds of millions of users lost access to it on June 12 and have not regained it. Anthropic's ongoing Pentagon litigation - the supply chain risk designation from earlier in 2026 - continues separately from the Mythos export control resolution.
Before the models were pulled, Mythos 5 had been available to trusted organizations involved in Anthropic's Project Glasswing, including infrastructure providers like Cisco and banks like JPMorgan Chase. Restoring access to that network of infrastructure defenders is the immediate priority. Restoring Fable 5 for general use is the next step - and the one with the largest business impact.
For enterprises that built workflows on Claude models before June 12, the message from this resolution is that access to Anthropic's most capable models is now a managed government program, not a standard enterprise software purchase. That reality will shape enterprise AI procurement decisions for years.
Cut Through the Noise
What did the US government allow Anthropic to do on June 26, 2026?
Commerce Secretary Howard Lutnick sent Anthropic a letter on June 26 permitting approximately 100 companies and federal agencies to access Claude Mythos 5, the company's most advanced cybersecurity AI model. The permission is subject to the government's right to modify the approved list at any time. Fable 5, the publicly accessible model disabled at the same time, remains unavailable pending further negotiations.
Why is Mythos 5 available to 100 partners but Fable 5 is not restored for general access?
Mythos 5 was always restricted to a small group of vetted organizations through Anthropic's Project Glasswing program, including critical infrastructure providers and financial institutions. Restoring its availability to that known, trusted group was a simpler security question than restoring Fable 5 to general global use. The government's primary concern was that Fable 5's guardrails could be bypassed, allowing general users access to Mythos-level cybersecurity capabilities.
What does Lutnick's reservation clause mean for Anthropic's model deployment?
Lutnick's letter stated he reserves the right to "reevaluate and adjust the scope of license requirements on the Covered Models, should circumstances change." This means the US Commerce Department has established itself as an ongoing gatekeeper for who can access Anthropic's most capable models - the government can expand or contract the approved partner list at its discretion, without Anthropic having autonomous control over its own deployment decisions.
How does Anthropic's situation compare to OpenAI's handling of government relations?
OpenAI voluntarily consulted with the government before launching its new GPT-5.6 models on June 26 and agreed to a managed rollout to trusted partners first. Anthropic received a forced shutdown, weeks of negotiations, and a partial restoration under government supervision. The contrast illustrates the practical value of proactive government engagement - OpenAI's relationship with the Trump administration, reinforced by its June 18 hire of Trump AI policy co-author Dean Ball, allowed it to navigate the same scrutiny environment without a forced outage.
