关键词: 生成式人工智能, 信息失实风险, 虚假信息犯罪, 水印标识, 标识消除行为

Abstract:

With the large-scale application of generative artificial intelligence， how to deal with the risk of inaccurate information it brings has become an important social issue. The evaluation results show that in the application scenario of generative artificial intelligence as the subject of false information fabrication， the current regulatory system for false information crimes cannot effectively respond due to the addition of risk behavior types. In the selection of governance plans， the limited autonomy of technology determines that the governance path that endows generative artificial intelligence with limited legal personality is not effective， and the regulatory focus should still be on the behavioral subjects behind generative artificial intelligence. Under the regulatory concept of balancing the protection of legal interests and the protection of freedom， a multidimensional co-governance criminal governance plan should be adopted. In the entire process of risk management， the automatic generation of information watermark identification is used as a link， adhering to the regulatory approach of prioritizing preexisting laws and criminal law protection， and forming a top-down linkage between different entities. In terms of generation dimension， information identification obligations should be imposed on service providers， and in the dimension of dissemination， identification elimination behaviors should be appropriately criminalized in order to tighten the criminal law net.

Key words: generative artificial intelligence, risk of information inaccuracy, false information crimes, watermark identification, identification elimination behaviors