A recent incident has highlighted the potential vulnerabilities of sensitive information, as a neural network model developed by Anthropic, called Mythos, was able to access nearly all classified resources of the US National Security Agency (NSA). According to Senator Mark Warner, the breach occurred in a matter of hours. This incident led to the White House demanding that access to advanced AI models, including Fable 5 and Mythos 5, be blocked for foreign users. As a result, Anthropic was forced to restrict access to the neural network. The Pentagon had previously taken steps to limit the use of Anthropic's AI models, with Defense Secretary Pete Higgs highlighting the potential risks associated with these advanced technologies. In a related development, the Pentagon utilized a chatbot called Grok, developed by xAI, a company owned by Elon Musk, in military strikes against Iran. The US government has since come out in support of xAI in a legal dispute related to air pollution allegations. The incident raises important questions about the technical architecture and system design decisions that can impact the security of sensitive information, and the practical engineering trade-offs that must be considered when developing and deploying advanced AI models.

Source: Lenta.RU