Private Algorithms And Public Interest: Overhauling The Trade Secrets Regime For Equitable Al Futures

The use of trade secrets to enclose the data undergirding artificial intelligence (AI) systems is a dimension that remains under-explored. This policy brief posits that the data enclosed in trade secrets by digital transnational corporations has the effect of stifling genuine innovation and makes Al systems non-transparent and unexplainable. While trade secret regimes are important for the functioning of innovative markets, they have tended to extend outwards and cover an increasing number of information goods of the nature of data in both commercial and non-commercial contexts. For instance, trade secret claims in the information-feeding recidivism algorithms have been used to deny requests by incarcerated individuals to understand why they were given a particular rating. The increasing prominence of Al in economic and social life compels an examination of the extent to which Al-related innovations should be protected under trade secret provisions. Trade secret protections are increasingly used to evade data or algorithm-sharing mandates in lieu of intellectual property protections where the latter are deliberately kept sparse for public welfare objectives. This policy brief examines the different impacts of trade secret regimes in the data and AI paradigm and offers forward-looking recommendations to ensure that trade secret protections do not end up creating monopolistic control over data, and that there is a transparent, inclusive, equitable, and accountable Al system. Key words: Al systems, data, trade secrets, intellectual property, digital transnational corporations