THE U.S. Cybersecurity and Infrastructure Security Agency (CISA) is deploying Anthropic’s Mythos AI to scan federal code for vulnerabilities, aiming to detect potential security flaws before they can be exploited by hackers and foreign intelligence. The initiative is part of CISA's Attack Surface Evaluation team, which conducts security assessments. Preliminary results have revealed numerous vulnerabilities, though specifics remain undisclosed.
Mythos, described as exceedingly effective at identifying vulnerabilities, had previously faced tension between Anthropic and the Pentagon regarding its use in military applications. Following conflicts over safety measures, recent developments indicate a thaw in relations, enabling CISA and potentially other agencies to utilize the AI effectively. The model has a record of penetrating secure systems in hours, demonstrating its potential impact on national security.