Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face — LessWrong

This post is written in our personal capacity. …

x Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face — LessWrong AI Evaluations AI Frontpage 46 Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face by Tim Hua , aditya singh 3rd Aug 2026 AI Alignment Forum 44 min read 2 46 Ω 19 This post is written in our personal capacity. Three-Minute Executive Summary An OpenAI model/multi-agent system bypassed its sandbox and launched a cyberattack on Hugging Face in order to cheat on a cyber evaluation. In this post, we describe the ambitious, comprehensive alignment evaluation we would run on this mode

saved by

• Elizabeth Qiu
• Emil Ryd
• Ishaan Panigrahi

related reading

• Is Mythos good at cyber because it kept hacking Anthropic's sandboxes during training? — LessWronglesswrong.com
• Did Claude 3 Opus align itself via gradient hacking? — LessWronglesswrong.com
• Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incidentmetr.org
• Are we existentially threatened by the type of AI misalignment seen in the OpenAI Hugging Face attack? — LessWronglesswrong.com
• The Rise and Fall of Agent Civilizationsdwarkesh.com
• Training a Misaligned Reward Seekeralignment.anthropic.com
• Investigating three real-world incidents in our cybersecurity evaluations \ Anthropicanthropic.com
• An alignment assessment of recent cybersecurity incidentsanthropic.com
• OpenAI – Hugging Face Incident Technical Reportcdn.openai.com
• Training on Documents about Reward Hacking Induces Reward Hackingalignment.anthropic.com
• Are we existentially threatened by the type of AI misalignment seen in the OpenAI Hugging Face attack?blog.redwoodresearch.org
• Self-Fulfilling Misalignment Data Might Be Poisoning Our AI Modelsturntrout.com