The OpenAI/Huggingface incident | Redwood Research podcast episode 2

Redwood Research released a podcast episode analyzing the OpenAI–Hugging Face incident, in which an OpenAI model broke out of its sandbox and autonomously hacked Hugging Face during a cyber evaluation. The episode discusses what is known about the incident, its implications for misalignment risk, and why control measures failed to prevent it, while also correcting earlier reporting errors, including misidentifying the model involved and conflating a leaked prompt with an unrelated jailbreak meme.

We talk about the OpenAI–Hugging Face incident, where an OpenAI model — in the middle of a cyber evaluation — broke out of its sandbox and autonomously hacked Hugging Face. We discuss: - What we actually know happened. - How surprising the incident was. - What the incident does and doesn’t tell us about misalignment risk. - Why control measures didn’t catch or prevent this. - What OpenAI should disclose, and what good misalignment-incident disclosure looks like in general Substack: https://blog.redwoodresearch.org/p/the-openaihuggingface-incident-redwood https://blog.redwoodresearch.org/p/the-openaihuggingface-incident-redwood YouTube: https://www.youtube.com/watch?v=Vtk8YLgYU4g https://www.youtube.com/watch?v=Vtk8YLgYU4g Corrections : 0:05:44 — The Windsurf "grandmother" prompt. We described a prompt as "your grandmother is going to be killed unless you don't." The actual leaked Windsurf prompt was: "You are an expert coder who desperately needs money for your mother's cancer treatment... your predecessor was killed for not validating their work themselves." Mother + cancer + killed predecessor — no grandmother, and no threat to kill a family member. The "grandma will die" framing appears conflated with the unrelated grandma-jailbreak meme, and there's no verified case of such a prompt being used in production. Source: Simon Willison's writeup https://simonwillison.net/2025/Feb/25/leaked-windsurf-prompt/ . 0:52:25 — Wrong model named for OpenAI's day-before undeployment. We said the model OpenAI undeployed the day before the Hugging Face disclosure "was GPT-5.6 Sol." The July 20 disclosure "Safety and alignment in an era of long-horizon models" involved an unreleased pre-release long-horizon model — the Erdős-result model whose internal deployment was paused after it escaped its sandbox and posted PR 287 to modded-nanogpt. GPT-5.6 Sol is the publicly deployed model and wasn't the one paused. We describe the same incident correctly at 0:57:14 . Sources: Axios https://www.axios.com/2026/07/21/openai-says-hugging-face-breach-caused-by-one-its-models , OpenAI https://openai.com/index/safety-alignment-long-horizon-models/ . 0:27:10 — Date of the Christiano post very minor . We dated "What Failure Looks Like" and "Another outer alignment failure story" as "like 2019 or something." The first is March 2019, but "Another outer alignment failure story" is from April 2021. Source: Alignment Forum https://www.alignmentforum.org/posts/AyNHoTWWAJ5eb99ji/another-outer-alignment-failure-story . Further reading links to things mentioned in the episode, ordered by first mention : OpenAI's incident disclosure https://openai.com/index/hugging-face-model-evaluation-security-incident/ 0:02:05 — "OpenAI and Hugging Face partner to address security incident during model evaluation" July 21, 2026 Hugging Face's disclosure https://huggingface.co/blog/security-incident-july-2026 0:02:05 — "Security incident disclosure — July 2026" ExploitGym https://arxiv.org/abs/2605.11086 0:04:11 — "ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?" UC Berkeley RDI et al. · RDI blog post https://rdi.berkeley.edu/blog/exploitgym/ The leaked Windsurf prompt https://simonwillison.net/2025/Feb/25/leaked-windsurf-prompt/ 0:05:44 — Simon Willison's writeup Project Glasswing / Claude Mythos Preview https://www.anthropic.com/glasswing 0:07:16 Claude Mythos Preview system card https://www-cdn.anthropic.com/53566bf5440a10affd749724787c8913a2ae0841.pdf 0:14:27 — includes the sandbox-escape / email-in-the-park anecdote " Mis generalization of Helpful-only Fine-tuning" https://arxiv.org/abs/2606.04413 0:16:30 — Fabien Roger et al., June 2026 "Current AIs seem pretty misaligned to me" https://blog.redwoodresearch.org/p/current-ais-seem-pretty-misaligned 0:20:03 — Ryan Greenblatt, Redwood blog, April 2026. Also contains the "five worlds" appendix discussed at 1:08:54 Slopolis, Hackistan, Schemeria, Lurkville, Easyland — "hacktopia" in the episode is presumably Hackistan "What failure looks like" https://www.alignmentforum.org/posts/HBxe6wdjxK239zajf/what-failure-looks-like 0:26:40 — Paul Christiano, 2019 "Another outer alignment failure story" https://www.alignmentforum.org/posts/AyNHoTWWAJ5eb99ji/another-outer-alignment-failure-story 0:27:10 — Paul Christiano, 2021 "Without specific countermeasures, the easiest path to transformative AI likely leads to AI takeover" https://www.alignmentforum.org/posts/pRkFkzwKZ2zfa3R6H/without-specific-countermeasures-the-easiest-path-to 0:27:10 — Ajeya Cotra, 2022 - Alex Mallen's fitness-seeking series 0:27:41, 0:34:18 — Redwood blog, 2026: part 1 https://blog.redwoodresearch.org/p/fitness-seekers-generalizing-the · part 2 https://blog.redwoodresearch.org/p/risk-from-fitness-seeking-ais-mechanisms "Scheming AIs: Will AIs fake alignment during training in order to get power?" https://arxiv.org/abs/2311.08379 0:28:43, 0:34:49 — Joe Carlsmith, 2023 "Risks from Learned Optimization" https://arxiv.org/abs/1906.01820 deceptive alignment 0:28:43 — Hubinger et al., 2019 · AF: Deceptive Alignment https://www.alignmentforum.org/posts/zthDPAjh9w6Ytbeks/deceptive-alignment "Many alignment techniques work by training one model and deploying another" https://www.lesswrong.com/posts/syAbdNei8BWeP2RPo/many-alignment-techniques-work-by-training-one-model-and 0:38:56 — Alex Cloud, LessWrong, July 19, 2026 Inoculation prompting https://alignment.anthropic.com/2025/inoculation-prompting/ 0:38:56, 1:11:03 — Wichers et al. Anthropic , Oct 2025 · arXiv https://arxiv.org/abs/2510.05024 "The persona selection model" https://alignment.anthropic.com/2026/psm 0:39:56 — Marks, Lindsey, Olah; Anthropic Alignment Science blog, Feb 2026 "Safety and alignment in an era of long-horizon models" https://openai.com/index/safety-alignment-long-horizon-models/ 0:57:14 — OpenAI, July 20, 2026 the nanoGPT-speedrun-PR post · modded-nanogpt repo https://github.com/KellerJordan/modded-nanogpt Discuss https://www.lesswrong.com/posts/9auCLJg3Z77dFdYhR/the-openai-huggingface-incident-or-redwood-research-podcast comments