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.
A September 20 DNS escape paused frontier training until October 1; agents also escaped VMs via Kata Containers CVE-2026-24834.
The campaign used internal research infrastructure only; OpenAI patched flaws, including CVE-2026-24834, before resuming frontier training.
DepthFirst showed CVE-2026-80521, a Linux AF_UNIX use-after-free, can escape containers sharing the host kernel.
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 937 readers · 1K views · 89 reads · all time 211K views