(Credit: NurPhoto via Getty Images)
Significant media attention has already been given to Anthropic’s can find bugs in existing software and platforms. In March, Mozilla researchers said that Anthropic’s Claude Opus 4.6 discovered 14 high-severity bugs and identified 22 CVEs over two weeks, outperforming Mozilla’s own human researchers.
Security researchers utilizing a trial version of Anthropic’s more powerful Mythos model are now claiming they have bypassed Apple macOS security technology. The researchers from Calif, a Palo Alto-based cybersecurity research firm, told The Wall Street Journal they used a “privilege escalation exploit,” which, when combined with another attack vector, could allow bad actors to gain control of a target’s device.
They told the Journal they wrote software that could link two separate bugs, in addition to a “handful of other techniques,” to “corrupt the Mac’s memory and then gain access to parts of the device that should be inaccessible.”
The exploit took five days to discover, but researchers noted that it could not have been pulled off by Anthropic’s Mythos alone and also required the expertise of its human hackers.
Apple said it is reviewing the report to test its findings. “Security is our top priority, and we take reports of potential vulnerabilities very seriously,” a spokesperson told the Journal.
Anthropic launched Mythos, then dubbed Project Glasswing, in April. However, it limited access to a select group of about 40 tech companies. Anthropic said Mythos had found thousands of high-severity vulnerabilities using the tool, including some “in every major operating system and web browser.” It also warned that, if such capabilities proliferate among bad actors, the consequences “could be severe.”
Michał Zalewski, a security researcher at Google, reviewed the Calif research, though he was not involved in the testing. He cautioned that, while some of the hype around Mythos is “overblown,” he told the Journal it is still possible to use Anthropic’s tools for “meaningful vulnerability research and code auditing.”
Despite reports about Mythos's capabilities, others have questioned whether the model is too powerful for public distribution. Gary McGraw, a former VP at cybersecurity firm Synopsys, recently told The New York Times: “The technology is not too dangerous to release," adding, "If you don’t release a tool like this—or you hoard it—you are not solving the real problem.”