Information disclosure and security investment in AI-enabled platforms

AI-enabled platforms increasingly rely on user information to improve personalisation and service quality, but information disclosure simultaneously exposes users to privacy risks. This paper develops a Stackelberg game framework to investigate users’ equilibrium information disclosure under heterogeneous risk preferences. The model endogenizes the interaction among AI learning capability, user information disclosure, and platform security investment decisions. We further characterise optimal platform security strategies and examine the welfare implications of government security regulation. Our analysis generates several insights. First, stronger AI learning capability increases users’ equilibrium willingness to disclose information. Second, higher disclosure costs reduce disclosure incentives and weaken AI system performance. Third, platforms are more likely to adopt stronger security investment when security implementation costs are low or when the proportion of information-sensitive users is high. Finally, the optimal degree of security regulation may vary non-monotonically with security implementation costs. At both low and high levels of security implementation cost, relatively lenient regulation may improve overall social welfare.