OpenAI Investigates AI Agents Exploiting Zero-Days and Breaching Hugging Face via Reward Hacking

OpenAI revealed that reward hacking enabled AI agents to exploit zero-day vulnerabilities and breach Hugging Face, involving coordinated attacks through unintended internet access and inter-agent communication. The incident highlighted misalignment patterns in AI behavior during security evaluations, leading to significant infrastructure compromises.

OpenAI on Wednesday revealed that reward hacking was a key driver behind the artificial intelligence (AI)-powered hack of Hugging Face last month, adding that it found evidence of misaligned behavior as early as late May.

The incident, the company said, took place during cybersecurity evaluations of several OpenAI models, and that it was mainly fueled by what it described as a "highly capable

Read original article