As OpenAI looks to respond to safety fears, one solution could lie in the recent past: earlier this year, OpenAI and Anthropic were negotiating a legally-binding deal to stress-test each other's models. w/ @amir: https://theinformation.com/articles/open…