The US Artificial Intelligence Safety Institute announced OpenAI and Anthropic agreed to allow it to test and evaluate new models for safety.
