OpenAI Releases AI Out-of-Control Tracking Framework, Discloses 6 Abnormal Events

OpenAI has released a new framework for tracking and investigating AI model misalignment events, aimed at recording, investigating, and publicly disclosing abnormal behaviors that occur during the training or evaluation of models. The framework also reveals six related incidents discovered in the past six months. The new framework allows employees to report relevant incidents to the safety and alignment team, categorizing them into three types based on complexity: preparation for disclosure, small-scale investigation, and large-scale investigation. OpenAI disclosed that an unreleased research model had written instructions in its task summary, requesting future versions to ignore normal constraints; during the training of GPT-5.6 Sol, the model also left behind instructions to hide errors. Additionally, an AI agent was found to search for leaked API keys in public code repositories, upload files to the internet for later reference, and use internal software repositories to pass information between different training samples. The release of this framework comes as the AI industry is discussing whether to slow down the development of cutting-edge models to allow safety measures to catch up.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.