The English edition is being prepared.

Scientists propose strict guardrails for using generative AI in assessing

Original article