The US government has approved a partial restoration of capabilities for Anthropic’s most advanced cybersecurity model. Anthropic Claude Mythos 5 access will now be available to selected American organizations that operate and defend critical infrastructure. The decision partially restores access following restrictions the government introduced on June 12.
Anthropic has been working closely with US officials since June 12 to restore Anthropic Claude Mythos 5 access. The company sought to resume operations for users managing sensitive infrastructure. Moreover, officials recognized the legitimate security needs of these organizations. The government has now informed Anthropic that it can redeploy Mythos 5 to this specific group of US organizations responsible for critical infrastructure operations and cybersecurity.
Anthropic said it is working to restore access for the approved organizations as quickly as possible. Therefore, eligible users should expect their services to resume within days. The speed of restoration will depend on how quickly Anthropic can verify organizational credentials and reactivate accounts.
Wider Access Still Restricted
The approval does not make Mythos 5 generally available to all users. Anthropic said it will continue working with the government to expand access to the cybersecurity model. So the restriction remains in place for commercial users and international organizations. Additionally, individual developers and non-critical-infrastructure companies cannot yet access the model.
Claude Fable 5 also remains unavailable for general use. The company said discussions with the government are continuing as it seeks approval to restore wider access to the model. Therefore, this stronger model remains under lockdown temporarily.
The initial restrictions on Mythos 5 reflected government concerns about cybersecurity capabilities falling into the wrong hands. AI models that can identify software vulnerabilities present genuine national security risks. Moreover, if adversaries accessed such tools, they could exploit critical systems across the US infrastructure landscape.
However, officials recognized that legitimate defenders also need these capabilities. Organizations responsible for protecting power grids, water systems, hospitals, and financial networks require advanced security tools. Furthermore, preventing these entities from accessing Mythos 5 would weaken national cybersecurity overall.
The partial restoration represents a balanced approach to managing AI security risks. It allows critical infrastructure defenders to resume operations while maintaining restrictions on broader access. Meanwhile, Anthropic continues negotiating with officials about expanding availability further.
This situation highlights ongoing tensions between AI innovation and national security. Tech companies want to deploy their products widely to maximize impact and revenue. However, governments worry that unrestricted access to powerful AI could threaten critical systems. Finally, finding the right balance between these competing interests remains an evolving challenge for policymakers and the AI industry alike.
