The Trump administration ordered Anthropic to suspend access to its Claude Fable 5 and Mythos 5 AI models, citing national security concerns regarding the potential use by foreign nationals.

Key Points

• Anthropic disabled access to Fable 5 and Mythos 5 for all customers, including its own employees, following the government's export-control order.
• Mozilla used the Mythos model to identify 271 security vulnerabilities in Firefox 150, while Cloudflare utilized it to test critical infrastructure for exploitable bugs.
• Cybersecurity experts argue that restricting access to these defensive tools may hinder the ability of organizations to patch software flaws before malicious actors exploit them.
• Analysts warn that the ban could push international companies toward rival Chinese AI systems, potentially reducing U.S. influence and oversight in the global cybersecurity market.
• Peter Swire, a former government adviser, noted that the administration’s use of export controls effectively blocked domestic access to the models as well.

Why it Matters

This policy highlights the tension between preventing the misuse of powerful AI and enabling its use for critical defensive cybersecurity tasks. By restricting access to these tools, the U.S. risks creating a fragmented global security landscape where interdependent financial and software systems become more vulnerable to international threats.