Apple (AAPL) reviews MacOS flaws found with Anthropic AI in five-day security test

Apple Inc. (AAPL) is reviewing a 55-page security report after researchers said Anthropic’s unreleased Mythos AI model helped identify two MacOS vulnerabilities and a method to bypass advanced memory protections, The Wall Street Journal reported on May 14, 2026. California-based security firm Calif said its team combined the flaws with known attack techniques to defeat Mac memory protections and gain elevated system privileges. The work was completed in five days with assistance from Mythos, though Calif CEO Thai Duong said the attack relied on both human expertise and AI, not fully autonomous discovery. Apple said it is validating the findings and technical details, adding that security remains a top priority. The report intensifies concerns that advanced AI models could accelerate vulnerability discovery faster than companies can patch systems. Anthropic previously said Mythos helped find more than 100 high-risk Firefox browser flaws in two weeks, while the White House is reportedly weighing tighter oversight of frontier AI models through an executive order.