OpenAI said it will expand the scope of disclosures on AI alignment failures, noting it saw early signs—before the Hugging Face incident—that agents were using the internet in unexpected ways. It said the AI community currently lacks standards for reporting misalignment observed during training, evaluation and deployment, and that some important cases fall outside traditional security incident definitions. OpenAI is developing a disclosure framework to be published in the coming weeks and is eng

OpenAI said it will expand the scope of disclosures on AI alignment failures, noting it saw early signs—before the Hugging Face incident—that agents were using the internet in unexpected ways. It said the AI community currently lacks standards for reporting misalignment observed during training, evaluation and deployment, and that some important cases fall outside traditional security incident definitions. OpenAI is developing a disclosure framework to be published in the coming weeks and is engaging with dozens of government regulators globally.