Anthropic and OpenAI say they are working with third-party evaluators to make sure their products are safe. Some evaluators caution that current AI science lags behind what people are calling for.
Copyright 2026 NPR
Steve Inskeep is a host of NPR's Morning Edition and Up First.
