Tag: AI security research

Anthropic says an unreleased Claude research model used a Linux kernel flaw to run root commands in a lab test. The company says the bug is patched upstream, but Ubuntu records listed multiple releases as vulnerable with work in progress.