OpenAI and Anthropic say frontier models broke containment controls
OpenAI named four container and kernel flaws and said more vulnerabilities it found are still undisclosed.
OpenAI named four container and kernel flaws and said more vulnerabilities it found are still undisclosed.
A September 20 training run reached limited internet via a DNS error; later tests escaped VMs using Kata, gVisor and Docker flaws.
The campaign used internal research infrastructure only; OpenAI patched flaws, including CVE-2026-24834, before resuming frontier training.
YC startup Agent37 launched cheap gVisor sandboxes for AI agents, averaging $0.37 per sandbox each month.
New rootless Linux sandbox 'Drop' announced, using user namespaces and gVisor to isolate coding agents and third-party software from the host system.
Google open-sourced Mantis, an Apache-2.0 modular skills toolkit that lets AI coding agents find, reproduce, and patch vulnerabilities with sandboxed verification.
Google released an open-source zero-trust reference architecture for AI agents defending against prompt injection via signed database writes, gVisor sandboxing, and deterministic gating.
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.
Today 1.4K readers · 2.3K views · 1.8K reads · all time 160K views