AI is owned by it’s owner

In the event that AI-generated inventions are patentable, there still remains the question of who would own these patents. An AI should not be able to own a patent, not only because AI cannot legally own property, but because it would be a bad idea. The AI would not be motivated by a patent, the AI would not obtain any direct benefit from ownership, and the AI would not be able to effectively enforce a patent. This presents a number of obvious options for patent ownership such as the AI’s owner, developer, or user. The developer, user, and owner may be the same person or persons, or they may be different entities.

Ownership rights to AI-generated inventions should vest in an AI’s owner because it would be most consistent with the way personal property (including both AI and patents) is treated. If you own a machine that produces property then you own that property - whether a loaf of bread or valuable data which you choose to protect as a trade secret (leaving aside more complex cases in which multiple parties are involved or you did not have rights to the machine’s input [whether baking soda or another party’s data]). AI owner patent ownership could be taken as a starting point, although parties should be able to contract around this default, and as AI-generated inventions become more common, negotiations over these inventions may become a standard part of contract negotiations. So long as property entitlements are clearly allocated, these parties can ultimately work out the most efficient allocation of rights between themselves. However, a default ownership rule is still necessary to minimize overall transaction costs.