New macOS Vulnerabilities Discovered Using Anthropic's Mythos AI

Severity: High (Score: 64.5)

Sources: M.Uk.Investing, www.wsj.com, 9To5Mac, Cybersecuritynews, Pymnts

Summary

Security researchers at Calif, a Palo Alto-based firm, have uncovered two vulnerabilities in macOS that could allow privilege escalation, enabling unauthorized access to restricted areas of Apple computers. The vulnerabilities were identified while testing an early version of Anthropic's Mythos AI model in April 2026. The exploit combines two bugs and various techniques to bypass Apple's advanced memory integrity protections. Apple is currently reviewing the findings and has stated that security is a top priority. The vulnerabilities could potentially be exploited in conjunction with other attack methods, posing a significant risk to users. Details of the vulnerabilities will be disclosed once Apple addresses the issues. The report from Calif is 55 pages long and was delivered directly to Apple in Cupertino. Key Points: • Researchers at Calif discovered two macOS vulnerabilities using Anthropic's Mythos AI. • The exploit allows privilege escalation, potentially enabling hackers to take control of affected Macs. • Apple is reviewing the findings and prioritizing fixes for the identified vulnerabilities.

Key Entities

• Zero-day Exploit (attack_type)
• Apple (company)
• CWE-269 - Improper Privilege Management (cwe)
• investing.com (domain)
• T1068 - Exploitation for Privilege Escalation (mitre_attack)
• IOS (platform)
• MacOS (platform)
• Bugmageddon (vulnerability)