Australia Works With Anthropic to Address Mythos AI Cybersecurity Risks

Gate News message, April 23 — Australia said it was working with Anthropic on possible cybersecurity flaws after the limited release of Mythos AI model raised concerns about critical infrastructure. The central banks of Australia and New Zealand were monitoring the release, while the government tracked emerging vulnerabilities.

Anthropist's Mythos model demonstrated significant exploit capabilities during early testing, successfully recreating software flaws and writing working exploits in more than 83% of cases on its first try. The model discovered a 27-year-old vulnerability in OpenBSD, a security-focused operating system, and could link four separate bugs to break through browser and operating system sandboxes. Experts warned that advanced coding and autonomous tools could accelerate sophisticated cyberattacks, particularly in banking where older, interconnected systems remain common.

The release is reshaping the cybersecurity industry's economics. As AI makes vulnerability discovery cheaper and easier to scale, the value proposition is shifting away from costly discovery work like penetration testing toward remediation and risk management. However, this transition pressures small volunteer teams maintaining critical open-source projects. Cyber insurers are expected to reprice risk quickly, likely adding exclusions for AI-discovered vulnerabilities left unfixed past specified deadlines.