US ban on Mythos is related to a jailbreak research by Amazon researchers (timesofindia.indiatimes.com)

The US government has mandated that Anthropic suspend access to its advanced AI models, Mythos 5 and Fable 5, due to national security concerns linked to a jailbreak research technique employed by Amazon researchers. This technique involved using specific prompts to prompt the model to disclose a limited number of known software vulnerabilities. Anthropic disagrees with the government's decision, asserting that the identified vulnerabilities were minor and had been publicly documented, suggesting that similar vulnerabilities could be uncovered by other publicly available models without requiring any bypass techniques. This ban is significant for the AI/ML community as it underscores the tension between innovation in AI research and national security interests. Anthropic's models, particularly Mythos 5, are designed to assist governments and companies in addressing software vulnerabilities, raising questions about the implications of restricting access to such advanced technologies. The broad application of the directive to encompass all users, including foreign nationals within the US, highlights the complexities surrounding AI security and the balance needed between protecting intellectual property and ensuring national security.