Strengthening patents won’t decrease trade secrets – and the link outweighs the turn
Clark Asay, 20 - Professor of Law, BYU Law School “ARTIFICIAL STUPIDITY”, 61 Wm. & Mary L. Rev. 1187, nexis uni //DH
It is also the case that in many AI contexts, parties simply prefer trade secrecy over patent protection. 395In such scenarios, it would thus seem that strengthening AI patents would have little to no effect on the parties' intellectual property choices. That preference for trade secrecy, in turn, is likely to continue to push the industry towards consolidation in many cases, regardless of the relative strength of AI patents. This is so for the reasons discussed above: trade secrecy tends to increase transaction costs relative to patent protection, meaning that vertical integration becomes the more economically efficient means of producing the AI solution, even if it falls short of being the most innovative one. 396
Indeed, this point is buttressed by the fact that important parts of AI systems are either not protectable under patent law (e.g., the training data) or only barely so because of the existing AI prior art. 397And as discussed above, some of those complementary assets are more feasible to access when the industry is consolidated. Hence, strengthening AI patents by expanding patent eligibility is not a panacea for the AI industry. Indeed, doing so may actively harm AI innovation by reviving an AI patent troll market. Instead, the real solution may lie outside of intellectual property rights altogether.