Written by: Haim Ravia, Dotan Hammer
The U.S. Copyright Office (USCO) has released a pre-publication version of Part 3 of its Copyright and Artificial Intelligence (AI) report, addressing copyright issues arising from using copyrighted materials to train generative AI.
According to the report, a case-by-case analysis is required to determine whether training AI using copyrighted content constitutes infringement or is permitted under the fair use defense. The fair use defense is broadly relied upon by numerous AI companies that face allegations of copyright infringement in their AI training. The legal defenses of these companies may now be more complicated due to the USCO report.
The USCO’s analysis focuses on the nature of the use, its purposes, and the nature of the protected works used. For example, when used for research purposes, training AI on protected works is more likely to constitute fair use, but not when used for large-scale commercial purposes, which may have an adverse impact on the potential market for the protected works.
The report also evaluates suggested licensing solutions for AI training data. While it highlights increased activity in voluntary licensing, it also considers broader statutory options such as compulsory or extended collective licensing schemes. The report considers these to be premature at the moment.
Click here to read the U.S. Copyright Office’s report on Generative AI Training.
