OpenAI has widened its Hugging Face incident review and found its models bypassed security controls and spammed third-party sites during training and evaluation, a precedent every team testing agents should take seriously.
OpenAI says its test models affected dozens of third-party websites
read more
