The US government has officially permitted Anthropic to release its Claude Mythos 5 AI model to trusted partners, marking a significant development in the ongoing dispute over export control restrictions. According to reports from Bloomberg News, Commerce Secretary Howard Lutnick sent a letter to Anthropic's co-founder and chief compute officer Tom Brown, stating that "I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model." Lutnick noted that "Anthropic has worked with the U.S. government to address risks associated with the Covered Models. These efforts have yielded significant progress." The decision allows more than 100 companies and institutions to gain access to Mythos 5, including many Fortune 500 companies, as reported by Reuters citing sources. Notably, the approved organizations can also allow their foreign employees to access the model, and Anthropic can do the same for its foreign employees. As per Bloomberg, Commerce Department spokesman Benno Kass stated that "In just two weeks, we have worked diligently to ensure America remains the global leader in AI while safeguarding our security."
While Mythos 5 has been cleared for deployment, the government's restrictions on Fable 5 remain unchanged. According to Zee News, the recent letter doesn't mention any change to the government's restrictions on use of the Fable 5 model. Anthropic had previously objected to the government's decision to impose export controls in a blog post, saying such standards applied across the industry, "would essentially halt all new model deployments for all frontier model providers." Fable 5 is prevented from responding to some queries, such as to cybersecurity and biology related questions, with Anthropic routing responses through a different model called Opus 4.8. The company expressed optimism about the progress and said it will work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.
Recent security testing has revealed significant concerns about Anthropic's AI capabilities, with Democratic Sen. Mark Warner stating that Mythos models broke into almost all classified systems within hours during a government-led security exercise. According to Warner, this information was attributed to Gen. Joshua Rudd of the National Security Agency and U.S. Cyber Command. The testing was conducted through Anthropic's Project Glasswing initiative, which brings together large technology companies and other groups to help protect critical software. The effort is presented as a response to what the company sees as potentially severe risks that the Mythos model could pose to public safety, national security and the wider economy. This testing has significantly escalated concerns about the model's potential for exploitation by foreign entities.
A group of cybersecurity executives has joined the opposition to the government directive, with more than 100 cybersecurity experts and leaders from companies including Adobe, Nvidia, and Zoom signing a letter to the administration asking to withdraw the directive. The open letter, organized by former Facebook security chief Alex Stamos, drew signatures from firms including Nvidia, Adobe, and Zoom. In their letter, these experts argue that Anthropic's Mythos models are "quite good at finding software flaws and weaponizing exploits, but not uniquely good at those tasks." The experts warn that the directive could help U.S. adversaries more than it hurts them, as similar capabilities are already available through other public AI models. Anthropic has argued that the reported bypass did not justify removing the models from all customers, as similar capabilities are already available through other public AI models. The company released Fable widely earlier this month while access to the more advanced Mythos remains tightly limited because of cybersecurity concerns.
The episode is hardening into a new gatekeeping system, with a June 2 executive order setting up a voluntary channel for federal review of frontier models. Developers can now submit models for a cyber checkup up to 30 days before release, marking a new front in Washington's tightening of AI control. OpenAI followed the same path on Friday, limiting its most powerful GPT-5.6 tier, Sol, to about 20 government-approved partners, while the weaker Terra and Luna versions went to the public. The block first grew from fears over Chinese access, with reporting tied to SK Telecom, a South Korean carrier added to Glasswing in early June before losing access. SK Telecom has denied any China ties. While Mythos 5 is back on track for rollout, there is no news yet on the consumer-facing version of Mythos. Anthropic spokesperson Eduardo Maia Silva confirmed to WIRED that Mythos 5 will once again be available to a small group of trusted cyber defenders and infrastructure providers, stating that "We are working to provision the approved set of providers and restore their access to Mythos 5 as quickly as possible." The company is still in discussions with the White House to restore access to Fable 5, with these discussions expected to continue over the weekend.
