Protection of original literary, musical, dramatic, visual, and architectural works of authorship under Title 17 of the United States Code.
Modern copyright law is confronting the fundamental boundary between expressive human authorship and non-expressive computational digestion. While 17 U.S.C. § 102 originally conceived of fixed tangible media authored by natural persons, the advent of high-dimensional generative models has shifted the battleground to whether intermediate vectorization, neural weight parameter checkpoints, and token embeddings constitute unauthorized derivative works under § 106(2) or protected transformative uses under § 107.
The legal characterization of high-dimensional neural network weights, embedding vectors, and diffusion model parameter checkpoints as potential derivative works.
Mass downloading and vector conversion of copyrighted text, imagery, and code for analytical ingestion where the end-user query output is non-substitutive.
Reimplementation of declaring code, function headers, and system interfaces to allow developer skill portability and ecosystem interoperability.
The constitutional and statutory rule requiring natural human intellectual conception for copyright authorship, denying registration to autonomous AI outputs.
The inter-circuit split between the Sixth Circuit (zero tolerance license rule) and Ninth Circuit (allowing de minimis sound recording sampling).
The U.S. Copyright Office continues to affirm its strict human conception standard, refusing copyright registration for outputs created wholly by generative AI while examining statutory licensing and transparency mandates for training corpora. European regimes enforce DSM Directive Article 4 reservation opt-outs.
