USA Blocks Anthropic's New AI Models

Artificial intelligence company Anthropic has been compelled by the U.S. government to disable access to two of its advanced AI models, citing national security reasons. This directive, issued on a Friday, prohibits all foreign nationals, including Anthropic's personnel, from utilizing the technology, forcing the company to cut off service to all its customers.

Anthropic stated that the specific concerns of the authorities were not clearly communicated. However, the government reportedly became aware of a method to bypass security measures on the "Fable 5" model, which could potentially facilitate hacking. Fable 5 is a restricted version of the more advanced "Mythos 5" model, which has been withheld from public release due to its significant capabilities in identifying software vulnerabilities.

The company disagrees with the government's risk assessment, asserting that Fable 5 does not provide hackers with unique tools. Anthropic warns that such restrictive measures could hinder the entire AI industry's development. This situation unfolds amidst an ongoing legal battle, where Anthropic previously refused the Trump administration's requests to employ its technology for mass surveillance and autonomous weapons, leading to the Pentagon terminating its contract with the company.

Claude Mythos, developed by Anthropic, is an AI model designed to code and understand programming languages. The company claims it excels at finding and exploiting vulnerabilities, surpassing most human capabilities in this area. The process involves the AI analyzing isolated program source code, hypothesizing potential vulnerabilities, testing these hypotheses, and then attempting to chain together multiple minor vulnerabilities to create a more potent attack, ultimately generating a bug report.

“

Such measures can halt the development of the entire industry.