A security breakthrough shows researchers using techniques inspired by AI model Mythos to discover and chain two bugs in Apple's macOS, enabling memory corruption and access to restricted system areas.

Calif, a Palo Alto-based security firm, conducted the work in April while testing an early Mythos Preview as part of Anthropic’s Project Glasswing, which grants select partners access for proactive vulnerability hunting.

Calif demonstrated a method that links two bugs and multiple techniques to corrupt memory and escalate privileges on macOS.

Full technical details will not be released publicly until Apple addresses the underlying issues; the report was delivered in person and emphasizes responsible disclosure.

The article notes ongoing communication with Apple and suggests more work to come on patch development and verification.

Anthropic is withholding general release of Mythos while sharing it with trusted partners, reflecting concerns about AI-powered offensive tools and accelerating defender-attacker dynamics in cybersecurity.

Apple is known for promptly addressing vulnerabilities via security updates, and the disclosed findings are expected to inform patches benefiting millions of Mac users.

The broader context references related Apple security stories and policy debates, illustrating ongoing industry focus on security and AI-assisted vulnerability research.

Experts note that AI-assisted cybersecurity research accelerates vulnerability discovery, challenging patch timing for companies.

The Wall Street Journal reported on the findings, with discussions signaling a potential wave of AI-assisted exploits influencing policy and national AI strategy.

The article references Mythos' prior secrecy and its involvement in a cybersecurity initiative with Apple, directing readers to additional coverage.

Wall Street Journal coverage notes researchers presented findings at Apple’s Cupertino headquarters, highlighting Mythos’ notable effectiveness in this context.