Commerce Department Clears Anthropic’s Mythos 5
The U.S. Commerce Department has authorized Anthropic to provide its Claude Mythos 5 artificial intelligence model to a select group of approximately 100 private companies and federal agencies. The decision, communicated through a formal letter, grants limited access to the model while keeping the company’s separate Fable 5 model under continued restriction, according to reports confirmed by CNBC.
Safeguards and Export Control Hurdles
This approval arrives after a period of intense federal scrutiny regarding the deployment of high-capability AI systems. Commerce Secretary Howard Lutnick wrote to Anthropic co-founder Tom Brown, stating that “appropriate safeguards are in place” to permit usage by trusted partners.
The government has not authorized the release of the Fable 5 model. Earlier this month, Anthropic suspended access to both systems to comply with a U.S. government export control directive. That order mandated that the company restrict access to these systems by foreign nationals, both domestically and abroad, citing national security concerns.
Leadership Shifts in Washington Negotiations
The negotiations mark a shift in the government’s relationship with leading AI developers. Reports indicate that Tom Brown has taken the lead in discussions with the Trump administration, replacing Anthropic CEO Dario Amodei in that role.
The standoff highlights a growing trend of federal oversight regarding AI models. Anthropic’s Fable 5 model was marketed as an advanced offering, utilizing new safeguards intended to mitigate risks in specific high-risk areas. The company faces persistent regulatory headwinds; the Department of Defense previously designated Anthropic as a “supply chain risk,” prompting a lawsuit from the company that remains ongoing.
OpenAI and the Trusted Partner Framework
Anthropic is not alone in managing these federal requirements. On the same day the Mythos 5 authorization was confirmed, OpenAI released three new models: GPT-5.6 Sol, Terra, and Luna. OpenAI confirmed it is complying with federal requests to initially limit the rollout of these models to a small group of trusted partners. While OpenAI intends to make these models generally available in the coming weeks, the company shared its development plans and capabilities previews with the government prior to the Friday launch.
Summary of Current Model Deployments
| Company | Model(s) | Status |
|---|---|---|
| Anthropic | Mythos 5 | Approved for ~100 trusted partners |
| Anthropic | Fable 5 | Access suspended |
| OpenAI | Sol, Terra, Luna | Limited rollout to trusted partners |
The Future of Federal Oversight
The current regulatory environment suggests that the “trusted partner” model is becoming a standard framework for the initial deployment of advanced AI in the United States. As federal agencies and private developers negotiate the balance between rapid innovation and national security, the ability of companies to satisfy government-mandated safeguards stands as the primary prerequisite for wider model distribution.
Related reading