while I generally agree that models with this level of cyber capability should not necessarily be released to everyone on day one, if a company or developer can meet compliance requirements, they should be eligible - and eligibility should not be based on nationality.
Your point is interesting! And when DeepSeek releases a Mythos-class model, will OpenAI and Anthropic apply the same policy with a Mythos-class model already on the market?