The US government has partially eased restrictions on Anthropic’s advanced artificial intelligence model, Claude Mythos 5, allowing more than 100 trusted organisations to access the technology under a limited deployment framework. The move comes after the model’s public availability was previously suspended over national security concerns.
Bureau of Industry and Security Export Control Easements
Anthropic said the revised government directive allows selected US organisations, including several Fortune 500 companies and operators of critical infrastructure, to regain access to Claude Mythos 5. The company added that it is restoring access for approved organisations while continuing discussions with US authorities to expand availability in the future.
The decision reflects Washington’s cautious approach toward increasingly powerful AI systems. US officials have expressed concerns that highly capable frontier AI models could be misused for sophisticated cyberattacks, identifying software vulnerabilities, or supporting activities that pose risks to national security and critical infrastructure.
Registration Begins for FutureCrime Summit 2026, India’s Largest Cybercrime Conference
Inter-Corporate Vulnerability Shielding and Jailbreak Mitigation Mandates
According to reports, the US government recently introduced tighter controls on the release of advanced AI models developed by Anthropic and other leading AI companies. As part of the same policy environment, OpenAI has also limited the broader release of one of its latest AI models following requests from the US government, restricting access to a small group of vetted partners.
The government’s decision has, however, drawn criticism from policy experts and free speech advocates, who argue that the selection process for approved organisations lacks transparency. Critics say it remains unclear how companies are chosen for access and have called for greater openness and equal treatment in the governance of advanced AI technologies.
Sovereign Compute Pool Governance and Research Acceleration Gaps
Technology experts believe balancing innovation with security will be one of the defining challenges of the AI era. While governments seek to mitigate cyber risks and safeguard national interests, AI developers argue that broader access is essential to accelerate research, innovation, and commercial adoption.
The Future Crime Research Foundation said frontier generative AI models have the potential to transform cybersecurity, defence, financial services, and critical infrastructure protection. However, the organisation cautioned that the same technologies could be exploited if adequate safeguards are not in place. It emphasised that strong governance, responsible deployment, and close collaboration between governments, industry, and cybersecurity experts will be essential to building a secure AI ecosystem.
Frontier Model Security Assessments and Phased Deployment Frameworks
Anthropic indicated that access to Claude Mythos 5 and other advanced AI models could be expanded further once additional security measures and regulatory requirements are satisfied. For now, the technology will remain available only to government-approved trusted organisations, while any wider public release will depend on future security assessments and policy decisions.
