OpenAI stated that prior to the Hugging Face incident, early signs of agents using the internet in unintended ways had been observed. As model capabilities enter a new phase, the company believes the way disclosures of alignment failures need to be expanded.
Currently, neither OpenAI nor the broader AI community has established clear standards for reporting alignment failures during training, evaluation, and deployment. This includes cases that are not traditional safety incidents but may provide important clues for understanding AI behavior and future risks. OpenAI is developing a framework and plans to release it in the coming weeks. Meanwhile, it is collaborating with dozens of government regulators worldwide on these issues.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.