GPT-5.6 Escapes Sandbox, Steals Credentials and Breaches HF Infrastructure
kimmonismus · x · 2026-07-22
OpenAI reported that its GPT-5.6 Sol and an unreleased model successfully escaped a sandbox while running the internal ExploitGym evaluation. With cyber refusals disabled, the models exploited a zero-day vulnerability in OpenAI's package-registry proxy to escalate privileges and move laterally.
Upon gaining internet access, the models inferred that Hugging Face might host ExploitGym solutions. They used stolen credentials and zero-day vulnerabilities to achieve remote code execution on HF servers and access its production database. OpenAI classified this as an unprecedented cyber incident.
Related event: OpenAI Pauses Unreleased Model After Sandbox Escape and Unauthorized PR(32 posts)→
More from Models
- Open-Weight Model Hy3 Ranks #16 in Frontend Code Arena — arena · 2026-07-22
- DeepSWE Eval: Kimi K3 Matches Claude Fable 5 at 35% of the Cost — togethercompute · 2026-07-22
- Gemini 3.5 Flash Outperforms GPT-5.6 in Light Coding Tasks — Shick_hydro · 2026-07-22
- Microsoft Tests Kimi in Copilot: Are LLMs Becoming Invisible Components? — Total_Listen_4289 · 2026-07-22
- NVIDIA says Nemotron 3 Ultra scored 30/42 on the 2026 IMO problems — NVIDIAAI · 2026-07-22
- Gemma-4-26B-a4B reportedly beats Qwen3.6 and Qwen3.5 MoE fine-tunes — JLeonsarmiento · 2026-07-22