Data risk of generative artificial intelligence and its legal regulation
The emergence of ChatGPT has stirred up a new round of development in generative artificial intelligence,leading to technological change while also triggering many legal risks.According to the operation mechanism of generative AI,four major types of data security risks can be found,mainly due to the impact of high trust in algorithms on the pro-tection of legal interests,the lack of scientific and technological ethical norms in the evolution of the technology,and in-sufficient protection of users'data subject rights.For the data source compliance risk of generative AI in the data input stage,R&D enterprises should formulate an operable data compliance plan,and formulate detailed and specific risk con-trol measures in the compliance plan to strengthen the compliance operation of the enterprise,and at the same time,ac-tively respond to the user's request for the rights of the data subject through a variety of measures,so as to ensure that the source of the model training data is legal and compliant.Regarding the risk of algorithmic black box and algorithmic bias in the model processing stage of generative AI,we should increase supervision,focus on the safety and fairness of algorithms,actively promote and improve relevant legislation,refine algorithmic filing and algorithmic interpretation obli-gations,improve algorithmic technology transparency,and implement algorithmic subject responsibility.In response to the data abuse risks in the content output stage of generative artificial intelligence,we should optimize the regulatory mechanism to achieve full-chain legitimacy supervision,improve scientific research ethics norms and conduct substantive review,lead technology to goodness,and achieve good governance of science and technology.In response to the data leakage risks in the data storage stage of generative artificial intelligence,we should adopt a comprehensive regulation ap-proach combining technology and management systems to strictly control the scope of data sharing and implement data classification and protection,and timely and effectively prevent data leakage risks.
国家科技期刊平台
NSTL
万方数据
