The U.S. government has approved the limited return of Anthropic’s Mythos 5 artificial intelligence model. The decision allows a selected group of trusted organizations to use the advanced AI system again after access was suspended earlier this month because of national security concerns.
The approval came after U.S. Commerce Secretary Howard Lutnick informed Anthropic that the government was satisfied with the safety measures the company had put in place. Officials concluded that the new safeguards would allow approved users to access the powerful cybersecurity model while reducing potential security risks.
Under the latest decision, access to Mythos 5 will be restored to about 100 organizations. These include government agencies, private companies, and operators of critical infrastructure. The model will mainly be used for defensive cybersecurity work, helping experts identify software weaknesses, strengthen computer systems, and improve protection against cyberattacks.
Anthropic confirmed the decision in a public statement. The company said the government had approved the redeployment of Mythos 5 to organizations responsible for operating and defending critical infrastructure across the United States. Anthropic added that it would continue working closely with federal officials as access is restored.
The company is also seeking permission to restore access to Fable 5, its latest consumer-focused AI model. According to people familiar with the discussions, Anthropic plans to continue talks with government officials as both sides work toward an agreement that balances public access with national security concerns.
Earlier this month, the Commerce Department ordered Anthropic to disable both Mythos 5 and Fable 5 by using export control authorities. Officials argued that the two advanced AI systems could pose security risks if accessed by unauthorized individuals or foreign nationals. The restrictions affected both government partners and commercial users.
Government officials were especially concerned that users might find ways to bypass the safety protections built into Fable 5. Although the model included stronger restrictions than Mythos 5, officials believed additional testing and review were necessary before allowing broader public access.
Export control rules also played a major role in the government’s decision. Because Anthropic employs foreign nationals and several partner organizations also include international staff, the restrictions required the company to suspend access completely while officials reviewed compliance with federal regulations.
Following the shutdown, Anthropic sent senior scientists and engineers to Washington to meet with experts from the Commerce Department and the Office of the National Cyber Director. The teams worked together to strengthen security measures and develop a framework that would allow trusted organizations to continue using the technology without creating unnecessary risks.
Before access was suspended, Mythos 5 had already demonstrated strong cybersecurity abilities. Early testing showed that the model could identify thousands of previously unknown software vulnerabilities and security bugs. Those capabilities made it valuable for organizations responsible for protecting important digital infrastructure, including financial institutions, technology companies, and public agencies.
The latest government decision came only hours after OpenAI announced the phased release of its new GPT-5.6 family of AI models. The company said it had worked with federal officials before launching the models and had shared a list of trusted partners as part of the review process. OpenAI also stated that it plans to cooperate with the U.S. government on a more permanent system for evaluating future AI releases.
The federal government has increased its attention on advanced artificial intelligence throughout this year. In June, President Donald Trump signed an executive order directing federal agencies to strengthen cybersecurity protections and create a formal process for testing powerful AI systems before they become widely available. That review system is still being developed.
The Anthropic Mythos 5 decision highlights the growing effort to balance rapid AI innovation with national security. As advanced AI systems become more powerful, governments and technology companies continue searching for ways to protect critical infrastructure while allowing trusted organizations to benefit from new tools that can improve cybersecurity and strengthen digital defenses.
