US deploys Anthropic AI for cyber defense
What's the story
The United States Cybersecurity and Infrastructure Security Agency (CISA) is using Anthropic's artificial intelligence (AI) model, Mythos, to detect vulnerabilities in government software. The move comes despite recent tensions between Anthropic and the White House over the company's AI safeguards and restrictions on model access. This initiative shows a growing confidence among US agencies in advanced AI systems for cyber defense.
AI deployment
Mythos scans code repositories for vulnerabilities
Mythos is being used to scan government code repositories for potential vulnerabilities that could be exploited by cybercriminals or foreign intelligence agencies. The task is being carried out by CISA's Attack Surface Evaluation team, a specialized unit tasked with penetration testing and broader security assessments across government systems. The AI-assisted reviews have already uncovered numerous software flaws, although details about their scale or severity remain undisclosed.
Corporate challenges
IPO plans and tensions with the US government
The deployment of Mythos comes at a critical juncture for Anthropic, which has confidentially filed for a US initial public offering. The company has had a rocky relationship with the US government, especially after it refused to remove safeguards preventing its AI models from being used for autonomous weapons or domestic surveillance. This disagreement escalated when the Pentagon labeled Anthropic as a supply-chain risk, usually reserved for foreign companies suspected of espionage.
Legal intervention
Improving relations and expanding Mythos's use
A federal judge blocked the Pentagon's designation of Anthropic in March, easing immediate restrictions on the company. Relations between Anthropic and the US government appear to have improved with the release of Mythos, an AI model tailored for cybersecurity tasks. The National Security Agency (NSA) started using Mythos as early as April, despite the earlier Pentagon designation.
Strategy concerns
Political scrutiny and access restrictions
Despite the positive reception of Mythos by NSA analysts, Anthropic's cybersecurity strategy has been under political scrutiny. When the company launched a public-facing version of Mythos called Fable, it included additional safeguards to prevent offensive cyber misuse. The White House then reportedly pressured the company to restrict access for foreign users, leading to a temporary global shutdown of the model. The ban was lifted earlier this month.