Days after restricting its Fable 5 and Mythos 5 AI models for non-Americans, the Trump administration has issued a clear ultimatum to Anthropic: If the company wants the government to lift the strict controls placed on its most advanced AI models, it must stop arguing and start fixing the software’s security flaws. According to a report by Wired, government officials made it clear that the company must actively patch “jailbreaking”- a technique where users employ specific text prompts to trick and AI into built in safety rules.

While Anthropic has repeatedly claimed the government’s concerns are overblown and the real-world risks are minimal, the White House appears to be no longer interested in debating.

National security threat

The administration’s tough approach was supported by a recent report conducted by the National Security Agency (NSA). In its report, the NSA came to a conclusion that the most tech-savvy individuals will be able to deactivate the guardtrails of Fable 5. This is worrisome because the guardsrails were designed d ..

In the absence of such digital flaws, the administration believes that AI would have been capable of assisting in such perilous areas as automated hacking, advanced chemistry, and biology. However, as the report reveals, the whole situation is viewed as Anthropic’s problem by the administration.

The officials have noted that both the National Security Agency and the new department Center for AI Standards and Innovation, established at the Commerce Department, do not have enough manpower or time to find all loopholes in the software products that come into the market.

Commenting on this:

Despite the fact that the demands of the White House are clear, independent cybersecurity experts note that the government may be asking for the mathematically impossible. More and more computer scientists believe that AI safety become a matter of concern for people.