External safety researchers at Hugging Face launched a transparency push on Sept. 12 targeting the way frontier labs develop artificial intelligence models. The effort, led by Thom Wolf, seeks a role in an "embedded evaluators" program announced by Anthropic CEO Dario Amodei, which grants third-party researchers permanent, employee-level access to its internal systems.

CEO Clement Delangue said the Open Alignment Initiative is necessary because AI safety cannot be solved by a small number of labs operating behind closed doors. The group aims to integrate with Anthropic's 3 part plan to slow AI development by allowing outsiders to verify safety measures and assess model alignment during training.