Security researchers use Anthropic’s Mythos AI model to crack Apple’s macOS

Fortune Tech: “Researchers with the Silicon Valley security firm Calif say they were able to use Anthropic’s Mythos AI model to corrupt an Apple Mac’s memory and gain access to parts of the machine that they shouldn’t. It’s called a “privilege escalation exploit,” and it can be used to seize control of a computer. The fact that the researchers were able to use Mythos to find two bugs to leverage in an operating system is one thing; the fact that they were able to do so in Apple’s legendarily restrictive macOS is quite another. “Memory corruption remains the most common vulnerability class everywhere, including iOS and macOS,” the researchers wrote. “In security, if you can’t fully prevent something, you accept the risk [and] mitigate it by making exploitation more expensive … Apple is smart and controls the full stack, so they pushed many of these defenses directly into hardware and made bypassing them significantly harder.” Apple told the Wall Street Journal that it’s working to validate Calif’s findings. But the broader takeaway—that Mythos “helped identify the bugs and assisted throughout exploit development,” as the researchers wrote—underscores concerns that the powerful AI model, in the wrong hands, represents an unprecedented risk to computer security.”