Game analysis of collaborative supervision mechanisms for privacy protection in generative artificial intelligence

The rapid development and application popularization of generative artificial intelligence (GAI) technology have led to privacy protection challenges, which have aroused widespread concern. It is of great significance to explore the collaborative supervision mechanisms for privacy protection in GAI and to improve the efficiency of supervision. Under the collaborative supervision mechanisms of government administrative supervision and enterprise self-regulation, an evolutionary game model consisting of GAI enterprises, users, and hackers is constructed, which is combined with the simulation method to examine the influence of different factors on privacy protection in GAI and its supervision. The results indicate that there are four stable equilibrium strategies in the collaborative supervision game system. For enterprises, the government’s reward and punishment measures as well as the users’ trust in enterprises assist them in raising the awareness of responsible innovation. Especially when the cost of compliance is high and enforcement is constrained, rewards are likely to be more effective than increasing punishments in encouraging enterprises to implement responsible innovation. For users, their sensitive information disclosure can be reduced by increasing their sense of privacy protection through education and publicity. For hackers, the efficiency of privacy supervision can be enhanced with higher punishments for illegal activities.