The FTC's position has been criticized by copyright experts who argue that AI training should not be considered a violation if it doesn't infringe on any particular copyright owner's rights. Despite this, the FTC continues to endorse the idea that all training data must be licensed and that "style mimicry" is a concern. The author criticizes the FTC for endorsing these points without considering their anti-competitive and problematic nature, and for failing to encourage more competition as per its stated charter.
Key takeaways:
- The FTC has overstepped its bounds by commenting on AI and copyright issues, arguing that fair use is anticompetitive and results in “unfair competition,” which is contrary to the purpose of fair use.
- If AI systems had to license all the data they train on, it would be anti-competitive as only a few of the largest tech companies could afford such a license, limiting AI innovation to those companies with deep pockets.
- Three well-known copyright experts criticized the FTC for being misguided on these issues, expressing concern about the FTC's suggestion that AI training might be a violation where it “diminishes the value of [a creator’s] existing or future works.”
- The FTC published a one-sided report on “AI and Creative Fields,” endorsing the untested idea that all training data must be licensed and arguing that “style mimicry” is a concern, which goes against the FTC’s stated charter to encourage more competition.