Artificial Intelligence Regulation, Minimum Viable Products, and Partitive Innovation

This Essay identifies entrepreneurs’ experimentation with minimum viable products as a means for proposed AI-specific regulation to constrain innovation in other markets. To that end, the Essay coins the term “partitive innovation” to describe a business’s perspective when it applies a domain-agnostic, highly generalizable technology to introduce a product to a particular market, with a likely result of eliciting overbroad domain-specific regulations that impair alternative innovative uses of the underlying technology. This process is unfolding with AI, as broadly constructed proposed regulation can restrict innovation in adjacent fields by shifting software MVPs’ mainly ex-post regulatory regime to one with recurring duties or ex-ante obligations. Consequently, the upfront cost of developing and altering those MVPs increases, thus effectively creating a barrier to entry in a market other than the one targeted by AI regulation. The Essay then considers real-world examples of proposed AI policy, California’s Assembly Bill 331 and the latest markup of the European Union’s AI Act, to illustrate the kinds of provisions that leverage AI products’ generality to expand policy influence into other domains. The Essay finally reviews a sampling of potential options to mitigate the regulatory spillover, including clear textual authority limits, grace periods based on enterprise scale, and monitoring the Food & Drug Administration’s opposite trend of using Predetermined Change Control Plans to reduce the number of ex ante reviews needed for medical AI devices