“Anthropic has committed not to develop or deploy models that reach a new AI Safety Level (ASL) until the corresponding safety measures are in place.”