Encryption, spyware, and now Mythos: History shows why cyber export control doesn’t work
TL;DR
TechCrunch frames the Mythos dispute as another attempt to control dangerous cyber software through export rules: the U.S. ordered Anthropic to restrict models like Mythos and Fable for foreign access. The article compares the case with PGP in the 1990s, when U.S. arms-export logic tried to slow encryption and instead helped trigger the Crypto Wars.
Nauti's Take
Export controls feel politically clean because they promise a visible line. Software has rarely respected that line.
PGP moved through books, spyware vendors shifted jurisdictions, and capabilities spread faster than licensing regimes. The Mythos case mostly shows that the U.
S. still lacks a convincing answer for dual-use AI: too much openness is risky, too much restriction weakens the vetted defenders who were supposed to use these systems responsibly.
Briefingshow
Mythos is not a normal chatbot story because Anthropic itself positioned the model as unusually strong at vulnerability discovery and cybersecurity work. That safety framing can now become a regulatory liability. For companies, access to frontier AI is turning into an export, compliance, and jurisdiction problem, not just a product choice.
Sources