The US has slightly eased restrictions on Anthropic’s Mythos 5, although this relaxation makes it even clearer how the focus remains on domestic use.
More than a hundred departments and companies in the US are once again allowed to work with Anthropic’s Mythos 5. The US government had previously imposed a ban on the use of the model by non-Americans. Because Anthropic could not determine who was behind a PC while using the LLM, the company felt compelled to take Mythos 5 and Fable 5 (the consumer variant) offline. In the new rules, the US government is dropping the focus on the user’s nationality.
This means that employees of more than a hundred approved organizations and enterprises can once again work with the powerful model, regardless of their nationality. For Anthropic itself, the relaxation also comes as a relief, as even its own employees were denied access to Mythos 5 without US citizenship.
The government chooses the customer
Access remains limited to so-called ‘trusted’ organizations. These are primarily large corporations, and exclusively American enterprises. In this way, the US continues to keep the latest AI technology within its borders. Whether this strategy is useful in the long term is doubtful. Just today, it emerged that a Chinese company has built an open model with capabilities comparable to Mythos 5.
Furthermore, the entire process continues to be characterized by a total lack of transparency. No one knows the criteria on which companies are selected or excluded. The US government can currently decide arbitrarily who gets access to an American AI model, without clear criteria.
Anthropic and OpenAI are also not fans of this approach. “Extensive safety testing is not a bad idea,” said OpenAI CEO Sam Altman. “But I don’t like the idea of the government being able to choose the customers.”
Reuters reports, citing an anonymous source, that the intention is to make Fable 5 available again in the near future. Under what conditions that would be, or when, remains unclear.