Great news for Anthropic: the US government has significantly eased restrictions on its most advanced AI model, Claude Mythos 5. This crucial development means that over 100 US organizations, spanning large corporations and government agencies, can now access the powerful AI. It marks a positive step forward after weeks of uncertainty for the cutting-edge AI developer.
The decision came in a letter from US Commerce Secretary Howard Lutnick to Anthropic’s cofounder and chief compute officer, Tom Brown. Lutnick confirmed that “appropriate safeguards are in place,” indicating that Anthropic’s dedicated work with the government to mitigate risks has paid off. This move signifies a cautious but clear endorsement of Anthropic’s efforts in AI safety and security.
Mythos 5 Gets a Green Light (Partially)
While the easing of restrictions is a significant win, it’s important to note that this isn’t a full green light for everything. The White House has permitted access to Mythos 5 for a select group of “trusted partners” but has held back on a broader, general rollout. Furthermore, the fate of Claude Fable 5, Anthropic’s consumer-facing version of Mythos, remains in limbo, still subject to the initial stringent safeguards.
Anthropic spokesperson Eduardo Maia Silva expressed the company’s delight, stating they “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.” The company is now working swiftly to restore access for these approved providers and remains committed to expanding access for Mythos 5 and making Fable 5 available for general use.
Crucially, the new directive also permits approved organizations to extend access to their foreign national employees, a reversal from the earlier prohibition. This change also applies to Anthropic’s own foreign national employees, simplifying internal operations and collaboration for the global team working on these advanced AI systems.
The Road to Reinstatement: What Happened?
The initial clampdown on Claude Mythos 5 and Fable 5 began roughly two weeks prior with an export control directive from the White House. This directive severely limited foreign nationals, even those residing and working in the United States, from accessing these sensitive AI models. It sent shockwaves through the AI community and raised questions about the future of AI development and access.
Concerns first arose under the Trump administration when Anthropic shared access to Mythos with a South Korean telecommunications firm, which officials suspected of having ties to China. Additionally, anxieties escalated when both Amazon and the NSA voiced worries that Claude Fable 5, despite its built-in safeguards, could potentially be “jailbroken” or bypassed. These combined fears prompted the urgent government intervention.
In response, Anthropic quickly mobilized its senior leadership. Key members from its cybersecurity and AI safety teams, including cofounder Tom Brown and public policy chief Sarah Heck, traveled to Washington, DC. Their mission was clear: engage directly with Trump administration officials and demonstrate the robust security and safety protocols in place for their frontier AI models.
Broader Ripples Across the AI Landscape
This saga with Anthropic highlights a growing trend of increased government scrutiny over advanced AI model releases. While getting Mythos 5 back online is a positive sign for Anthropic, the entire episode has ignited a larger discussion about the future direction of US AI policy. Many are now questioning the extent to which the Trump administration intends to control the rollout of new, powerful AI technologies.
The impact is already being felt across the industry. Just recently, OpenAI announced it would be delaying the release of its highly anticipated GPT 5.6 models, a direct response to a request from the Trump administration. This incident underscores a significant shift where AI developers might now require explicit governmental approval before launching their innovations.
Dean Ball, a former White House AI adviser and head of strategic futures at OpenAI, articulated this new reality in a recent blog post. He observed that Anthropic’s experience has effectively signaled to frontier AI model developers that they “need an explicit green light from the government now,” fundamentally altering the landscape for AI innovation and deployment.
For Anthropic, these government interactions haven’t been without significant costs. The company previously sued the Trump administration earlier this year over a supply chain risk designation, following its efforts to set clear boundaries on how military contractors could utilize its AI models. Furthermore, the June 12 export directive caused considerable concern among Anthropic investors, who spent the subsequent weekend evaluating the potential long-term implications for the company’s future and stability.
Source: Wired – AI
