The quandary of how to balance innovation and risk has taken on new urgency since large language model tools hit the public late last year. It's an issue Anthropic AI has grappled with since its founders left OpenAI in 2021 to start their own company focused on building an "honest, helpful and harmless" artificial intelligence system.

General counsel Brian Israel, who joined Anthropic in October, said risk calibration is a core part of his role.