The U.S. government lifted a block on Anthropic’s Claude Mythos 5 AI model on Friday, allowing more than 100 U.S. institutions and government agencies to access the technology. Commerce Secretary Howard Lutnick authorized the move after Anthropic committed to new safety protocols and export standards following recent security concerns.
Why did the U.S. government block Claude Mythos 5?
The administration imposed export controls on the Mythos model two weeks ago. This decision followed warnings from Amazon and other companies that the AI could be “jailbroken” for malicious purposes. The resulting block effectively shut down access to both the Mythos 5 model and its smaller counterpart, Fable 5.
The confrontation between the Trump Administration and Anthropic represented a significant tension between national security interests and the growth of private AI companies. The government’s primary concern centered on the potential for bad actors to bypass safety guardrails to use the model for unauthorized or harmful activities.
“Jailbreaking” in AI refers to the process of using specific prompts to bypass a model’s built-in safety filters, potentially forcing the AI to generate restricted or harmful content.
How will Anthropic’s new access work?
Commerce Secretary Howard Lutnick sent a letter to Anthropic’s chief compute officer, Tom Brown, stating that “appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model.” Lutnick cited “significant progress” in daily negotiations between the government and Anthropic since the block began.
Under the new arrangement, a license is no longer required to export, reexport, or transfer the Claude Mythos 5 model to specific entities. According to the letter, these entities include those identified in “Annex A” and their foreign national employees, as well as Anthropic’s own foreign national employees.
Anthropic has committed to working with the U.S. government to establish specific protocols, standards, and release procedures for its future models. This agreement marks a de-escalation in the regulatory friction between the company and federal authorities.
How does this compare to OpenAI’s latest release?
The lifting of the Anthropic block coincides with a major release from its primary competitor. On the same day the government cleared Mythos 5, OpenAI released its latest model, GPT-5, to a short list of government-approved partners.
| Feature | Anthropic (Claude Mythos 5) | OpenAI (GPT-5) |
|---|---|---|
| Access Scale | 100+ US institutions/agencies | Short list of approved partners |
| Regulatory Status | License waived for Annex A | Government-approved partners only |
While Anthropic is moving toward a broader institutional rollout, OpenAI’s GPT-5 remains restricted to a more exclusive group of partners, according to reporting from Axios.
What is the status of Fable 5?
The Commerce Department’s letter did not mention Fable 5, which is a less powerful version of the Mythos model. Fable 5 was previously the most widely available high-power consumer AI model before the export controls were implemented.
People close to the negotiations indicated that discussions are moving toward the release of Fable 5 as well. However, officials have not yet provided a specific timeline for when this model will be available to the public or broader partner groups.
When integrating new AI models into institutional workflows, always verify if the model’s deployment falls under specific export control annexes or federal “trusted partner” designations to ensure regulatory compliance.
Frequently Asked Questions
Which institutions can use Claude Mythos 5?
Access is currently limited to more than 100 U.S. institutions, including major corporations and government agencies identified in “Annex A” of the Commerce Department’s recent letter.
Why was the model blocked by the government?
The model was subject to export controls following warnings from companies like Amazon that the AI could be “jailbroken” to perform malicious tasks.
Is GPT-5 available to the general public?
No. According to current reports, OpenAI has released GPT-5 only to a limited list of government-approved partners.
Stay updated on AI regulation and industry shifts. Subscribe to our newsletter or leave a comment below with your thoughts on government oversight of AI models.
