A report from multiple outlets says researchers used Anthropic’s AI model “Mythos” to identify and connect macOS security weaknesses that Apple is currently reviewing. The work describes how the AI was able to link two minor software flaws, enabling a more serious outcome than either issue alone. According to the reporting, the linked vulnerabilities can allow an attacker to corrupt memory and potentially access sensitive data on a targeted Mac. The reports further state that, in some scenarios, chaining the flaws could lead to broader control of a victim’s computer, though details of exploitability are not fully established in the coverage. Anthropic’s Mythos is described as a model that has been kept relatively restricted due to concerns about misuse, but the findings presented by researchers are framed as security research rather than an open, public exploit release. Apple says it is reviewing the findings. Overall, the reporting highlights how AI-assisted techniques can accelerate vulnerability discovery and vulnerability chaining, potentially affecting real-world security risk assessments.