The U.S. government has started preparations to introduce artificial intelligence (AI) models with strong enough performance to transform ordinary hackers into special forces agents. Supply of the models to the private sector is strictly limited due to concerns over security threats.
According to Bloomberg News in the U.S., Gregory Babacha, head of federal informatization at the White House Office of Budget Management (OMB), recently said in an email to each department that he is working on security measures to allow government agencies to use Anthropic’s AI model Mythos. “Before distributing a revised version of the model to federal agencies, we are working closely with companies and intelligence agencies to ensure proper protection and guardrails,” said CIO Babacha.
The “Mythos,” which is the target of this introduction, is a model that has been determined to be a fatal risk to national security even within Anthropic. This is because internal tests showed overwhelming performance, such as finding key system defects that only the world’s best hackers could find in an instant. Defense experts have warned that hackers with the model in their hands will undergo “a change in the level of transformation from ordinary soldiers to special forces personnel.”
It should be noted that the Donald Trump administration has been strongly checking Anthropic. Earlier this year, the U.S. Department of Defense defined Anthropic as a “supply chain threat” that applies only to foreign hostile forces, citing AI security concerns. Anthropic filed a lawsuit against the government’s ban on use, saying it is expected to lose billions of dollars, and last month, it drew an order from the court to suspend the ban.
After the court ruling, the government’s mood has changed dramatically. The Treasury Department is requesting the right to use Mythos to find out the software that is already in operation, and major security agencies such as the Defense Ministry, the Commerce Department, and the Department of Homeland Security are also confirmed to be included in the list.
Anthropic is restricting the use of Mythos to only a small number of tech companies and financial companies. It is due to concerns that hackers could use the tool to steal data or destroy networks. However, the U.S. government plans to strengthen its cyber defense capabilities by allowing federal agencies to grab powerful AI tools ahead of the private sector through this measure.
Treasury Secretary Scott Bessent and Federal Reserve Chairman Jerome Powell also recently met with Wall Street leaders in Washington to recommend using the model to identify vulnerabilities in the financial system.
SAM KIM
US ASIA JOURNAL
