Artificial intelligence (“AI”) relies on the use of large datasets to train AI algorithms. Access to such data is therefore a critical resource, the lack of which may create barriers to entry for both AI startups and established firms developing AI technologies. We describe how existing and new competition policies may be appropriate for addressing these barriers to entry. We focus particular attention on how policies differ with respect to whether they encourage competition for the market or in the market. Potential new competition policies include temporary data monopolies, data portability, and the use of blockchain and trusted third parties.