US Copyright Office found AI companies breach copyright. Its boss was fired

Role of the Copyright Office and the Firing

  • Several comments clarify the Office’s mandate: to study copyright issues and advise Congress, not to decide cases; courts will ultimately determine legality.
  • The Part 3 report is framed as a response to congressional interest, but some see it as largely repeating rights‑holder complaints with thin reasoning.
  • The firing of the Register is widely interpreted as political: punishing an interpretation unfriendly to large AI firms, though details remain unclear.

Is AI Training a Copyright Violation?

  • One camp argues training on copyrighted works without permission is “obviously illegal,” especially when sources were pirated datasets (e.g., torrenting ebooks) or terms of use were ignored.
  • Others say current law targets copying and distribution, not “reading” or analysis; they analogize training to a human reading many books then writing something new, and emphasize that fair use is decided case‑by‑case.
  • Clear agreement that output which reproduces works verbatim (or nearly) is infringement, regardless of AI vs human. Disagreement is about whether training itself is infringement and whether models’ weights “contain” protected works.

Fair Use, Plagiarism, and Human Analogy

  • Repeated insistence that plagiarism and copyright are distinct: plagiarism is about attribution and integrity; copyright is about economic control and specific exclusive rights.
  • Debate over analogies: “perfect‑recall savant” vs. lossy learner; AI vs search index vs compression algorithm.
  • Some argue the key test should be whether outputs substitute for or harm the market for originals (books, music, journalism, code), not metaphysical questions about “inspiration.”

Economic and Ethical Concerns

  • Strong resentment that individuals were heavily punished for small‑scale piracy while tech giants mass‑copied books, music, and code with little consequence.
  • Critics highlight lobbying to entrench AI training as fair use, selective licensing deals (e.g., with major publishers), and lack of sanctions for large‑scale piracy as evidence of regulatory capture.
  • Others argue that blocking training in the US will just shift advantage to foreign firms; opponents reply that this is a “race to the bottom” justification.

Rethinking Copyright and Power Dynamics

  • Wide range of reform proposals: from abolition of copyright, to short fixed terms (e.g., 20 years), to “lifetime + floor,” to compulsory licensing schemes.
  • Ongoing tension between seeing IP as a necessary incentive for creators vs. a state‑granted monopoly now weaponized by corporations.
  • Several note a cultural shift: early internet pro‑piracy attitudes versus today’s strong defense of creators when the infringer is big tech rather than individuals.