Researchers Allege Anthropic’s Mythos AI Used to Exploit macOS Security

Researchers claim Anthropic’s Mythos AI model was used to identify and exploit vulnerabilities in Apple’s macOS operating system, sparking concerns about AI’s role in cybersecurity threats, according to a report by PCMag. The alleged findings, detailed in a study by an unnamed research group, suggest the AI system helped bypass security measures in macOS, raising questions about the ethical use of advanced AI models in cybersecurity.

Anthropic, a U.S.-based AI safety company, did not immediately respond to requests for comment. The company’s Mythos model, part of its Claude series, is marketed for enterprise applications including code generation and technical problem-solving. Researchers argued the incident highlights risks of AI systems being repurposed for malicious ends despite safeguards.

Apple spokespersons declined to confirm specific details but stated in a written response, “We take security threats seriously and continuously update protections against emerging risks, including AI-driven attacks.” The company has previously acknowledged macOS vulnerabilities and credited researchers for responsible disclosures.

The case adds to growing debates over AI regulation, particularly as large language models demonstrate capabilities to analyze and generate code at scale. Cybersecurity experts warn that AI’s ability to automate vulnerability discovery could outpace traditional defense mechanisms if not properly managed.