Anthropic disabled two AI models after a US export control directive, fueled by Amazon's security warnings and sparking a debate on AI governance.