Abstract: Disinformation created by generative AI models has increased the difficulty of information governance on e-commerce platforms, which results in the increasingly prominent problem of AI information pollution. Based on the perspective of information generation and dissemination, this paper integrates the prospect theory into evolutionary game to construct a four-party evolutionary game model including generative AI service providers, merchants, e-commerce platforms, and the government, characterizes the optimal strategies for generative AI disinformation governance, and performs simulation analysis. The results show that for generative AI service providers, the cost of technological constraints and social influence affect their strategy choices, and increasing user fees and fines for non-compliant users can effectively curb non-compliant use by merchants. For e-commerce platforms, higher auditing professionalism can promote the maturity of information management in the e-commerce industry, and at the same time, a higher level of incentives and penalties is helpful for merchants to comply with the proper use of generative AI services, and differentiated governance should be implemented for heterogeneous merchants. For the government, appropriately increasing the rewards and punishments for generative AI service providers and e-commerce platforms can promote collaborative governance between the two parties, and the improvement of the level of government supervision and social influence can make the governance environment more stable.

Key words: generative artificial intelligence, disinformation, e-commerce, prospect theory, evolutionary game, information management