Tech news in 3 minutes
From PGP to Mythos: a brief history of export controls that didn’t stop anyone
Last Friday, the White House ordered Anthropic to restrict exports of its powerful AI models Fable and Mythos, citing unspecified national security concerns. Anthropic quickly disabled both models, which have been unavailable for a week. This marks the first real test of using export controls to contain frontier AI. Anthropic had marketed Mythos as a potential cyber threat, limiting access to about 150 vetted companies and government organizations. The ban followed two events: Anthropic gave a South Korean telecom (reportedly SK Telecom, which denied ties to China) access to Mythos, alarming US officials; and Amazon CEO Andy Jassy alerted the administration after researchers found a way around Fable 5’s safeguards (Anthropic disputes this as a narrow, patched issue). The Commerce Department issued an export control directive, and Anthropic complied within 90 minutes. The article draws parallels to past export control efforts. In the 1990s, the US government investigated PGP creator Phil Zimmermann for violating arms export controls, but he published the source code as a book, igniting the "Crypto Wars" and paving the way for widespread encryption. Later, the Wassenaar Arrangement aimed to limit spyware exports but had weaknesses: non-signatory countries like Israel and lax enforcement in Europe allowed spyware to reach authoritarian regimes. Some successes include Germany-based FinFisher shutting down after a multi-year investigation. The current standoff remains unresolved. The administration might lift restrictions to keep US AI competitive, acknowledging that foreign labs may reach similar capabilities anyway. Alternatively, US AI companies may need government approval for foreign customers, hurting profits. Given historical failures, government-mandated export controls are unlikely to stop malicious abuse of powerful dual-use technologies.