Anthropic-linked researchers detail SharedRoot, a sandbox escape that can expose a host filesystem
EdenEmarco177 · x · 2026-07-24
Researchers said they found and reported multiple sandbox-escape vulnerabilities to Anthropic, and are now highlighting one called SharedRoot.
- The team argues AI-assisted kernel bug-finding is making such attacks more industrialized.
- Their claim: sandboxes are always effectively one N-day behind, so containment cannot rely on a clean guest Linux kernel.
- SharedRoot allegedly escapes the Cowork VM isolation layer and can grant unauthorized access to a user's computer, including the entire contents of the exposed filesystem.
More from coding & agent
- Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
- Dev argues Claude burns far more tokens than GPT models for the same coding work — haider1 · 2026-10-27
- Stop building agents that can't change models: stable contracts, external state, real evals — rseroter · 2026-10-27
- Cheap models can cost more in agent harnesses than frontier ones, researcher warns — vasuman · 2026-10-27
- Kanzo: A 5k-Line Rust Kanban Binary With MCP for Claude Code and Codex — allmudi · 2026-10-27
- 'A human approves everything' is a trap: approval fatigue and how to actually design safe agents — Efistoffeles · 2026-10-27