Encryption, spyware, and now Mythos: History shows why cyber export control doesn’t work
TL;DR
TechCrunch asks whether U.S. export controls on Anthropic models like Mythos can work better than earlier restrictions on encryption and spyware. The trigger was a White House intervention ordering Anthropic to restrict Fable and Mythos for users outside the U.S. and foreign nationals inside the country. The article connects PGP in the 1990s, the Wassenaar Arrangement, and modern spyware vendors that often route around rules or relocate to softer jurisdictions.
Nauti's Take
The instinct is understandable: if a model can perform dangerous cyber tasks, policymakers want to shut the tap. History says code rarely waits politely at the border.
A better path is access control, monitoring, liability, abuse response, and real international coordination. Pure export logic sounds tough, but it can become security theater with expensive collateral damage.
Briefingshow
Mythos is not a normal SaaS export case. It is a test of how governments handle dual-use AI that can help defenders and attackers. If the U.
S. treats every powerful cyber assistant like a weapons export, global AI products may need political clearance, which could reduce some risks while slowing research, defense work, and competitiveness.
Sources