Last Friday, the White House ordered Anthropic to halt the export of its AI models Fable and Mythos to non-U.S. entities, citing unspecified national security concerns, leading Anthropic to withdraw both models from all users.
Initial Trigger and Response
The U.S. government's move marks a significant test of export controls on advanced AI, paralleling previous attempts to regulate encryption and spyware. After Anthropic launched Mythos in April, it was promoted as a potentially disruptive tool, restricted to around 150 vetted organizations for security purposes.
The ban was reportedly triggered by two events: Anthropic's Mythos access granted to a South Korean telecom, speculated to be SK Telecom, raised alarms due to alleged China ties, which the company denies. Additionally, Amazon CEO Andy Jassy informed the administration of a security workaround in Fable 5, which Anthropic disputes as a minor patched issue.
Export Control History
Historically, export control efforts have faced challenges. In the 1990s, the U.S. attempted to curb encryption technologies like Pretty Good Privacy (PGP) by pursuing its creator, Phil Zimmermann, under arms export controls. Zimmermann countered by publishing PGP's source code as a book, ultimately leading to the widespread adoption of end-to-end encryption.
In the 2010s, spyware discoveries led to the expansion of the Wassenaar Arrangement, aiming to regulate dual-use technologies. However, the agreement's efficacy is limited by non-signatory countries and discretionary application by member countries, as seen with Italy's leniency towards Hacking Team.
Current Challenges and Implications
Despite some successes, like the shutdown of FinFisher following legal actions, export controls have struggled against relocation tactics by companies like Intellexa. The ongoing Anthropic impasse suggests that U.S. restrictions may not prevent the global development of similar AI capabilities.
If the U.S. government lifts the restrictions to support American AI competitiveness, it may tacitly acknowledge the inevitability of similar advancements abroad. Conversely, maintaining strict controls could impose regulatory burdens on domestic AI firms, impacting their global market reach.
Conclusion
The history of cyber export controls suggests that government mandates alone may not effectively prevent the misuse of dual-use technologies. The outcome of the current situation with Anthropic will likely influence future policy approaches toward AI and cyber technologies.
Source: https://techcrunch.com/2026/06/19/encryption-spyware-and-now-mythos-history-shows-why-cyber-export-control-doesnt-work/
