We study a platform's incentives to delist IP-infringing products and the effects of holding the platform liable for the presence of such products on innovation and consumer welfare. For a given number of buyers, platform liability increases innovation by reducing the competitive pressure faced by innovators. However, there can be a misalignment of interests between innovators and buyers. Furthermore, platform liability can have unintended consequences, which overturn the intended effect on innovation. Platform liability tends to increase (decrease) innovation and consumer welfare if the elasticity of participation of innovators is high (low) and that of buyers is low (high).

We study within-device competition between integrated (e.g., Google Play or Apple's App Store) and third-party app stores. We find that competition does not lead to lower commission fees. The reason is that ad valorem fee setting implies a negative pass through of fees to app prices (Edgeworth paradox), which induces developers to adopt price parity across stores (without being required to do so by the integrated platform). As a result, consumers have no incentives to install third-party stores, and the integrated store becomes a natural monopoly. Eliminating default advantages through choice screens is unlikely to be effective. Allowing developers to steer consumers towards alternative subscription methods lowers commission fees and increases the number of apps, but also increases app prices, which may decrease consumer surplus. 

We study the strategic behavior and optimal regulation of generative AI firms operating in markets with partial user unawareness of AI-related risks. We consider a model in which a monopolistic firm chooses its business model (subscription, freemium or free), pricing structure, and investment in algorithmic quality, and users' queries to the AI algorithm generate training data that can be used to improve the algorithm. We find that business model choice hinges on the potential to monetize algorithmic improvements in complementary markets and on the level of user awareness. Greater user awareness favors freemium or free models over the subscription one. Unregulated AI services may lead to suboptimal outcomes, with either excessive or insufficient investment in algorithmic quality, insufficient provision of algorithmic services, and low transparency of algorithms. We investigate regulatory interventions, such as transparency mandates and risk-management requirements, and show that these policies can mitigate user harm and align firm incentives with social welfare. We find that while full transparency is generally beneficial, in some cases it may be socially optimal to allow for full or partial transparency,