The English edition is being prepared.Scientists propose strict guardrails for using generative AI in assessingOriginal article