Human-Content-to-Machine-Data_Final - Flipbook - Page 23
Using contracts to govern what copyright does not also poses ethical quandaries, because it
could disrupt the balance struck between free expression and the rights of authors.90 In
cases where copyright does not provide the right to control a given use of a work, it likely
reüects a legislated compromise between the interests of creators and the public.
Permissionless reuse of copyrighted works plays an important role in the preservation of free
expression.91 This is why CC licenses overlay onto the acts copyright restricts—and not onto
acts protected under exceptions and limitations to copyright. This has been a fundamental
principle of CC licensing.
CC9s approach to forging a new social contract for machine reuse is therefore aimed ûrst and
foremost at changing norms.
The Need for a Nuanced Way to Express Preferences
This issue is often framed as a binary—either works should never be used by large AI models
without permission, or they should always be allowed.
However, as with the application of the law, the reality of people9s notions about what is fair
and prosocial when it comes to sharing and reuse of content are more complex. For example,
many African natural language processing experts continue to believe in openness and the
use of permissive licenses, while feeling that open licensing alone does not respond to their
concerns of extractive technology development.92 There are different feelings within
communities, too. Some, but not all, members of the Wikipedia community feel the
unrestricted crawling and use of data allows AI companies to unfairly exploit the web.93
At present, the opt-out approach for copyrighted works in the EU lacks nuance.94 We9re
concerned that bluntness could help bring about the lose-lose scenario of entrenching the
90
In some jurisdictions, contracts that attempt to control copyrighted works in ways that go beyond
the protections afforded by copyright could also face legal obstacles. See e.g., the discussion on
copyright misuse in The Mirage of Artificial Intelligence Terms of Use Restrictions by Peter
Henderson and Mark Lemley.
91
Netanel, N. (1996, November). Copyright and a Democratic Civil Society. The Yale Law Journal.
92
Okorie, C. & Marivate, V. (2024, April 30). How African NLP Experts Are Navigating the Challenges
of Copyright, Innovation, and Access. Carnegie Endowment for International Peace.
https://carnegieendowment.org/research/2024/04/how-african-nlp-experts-are-navigating-the-challenge
s-of-copyright-innovation-and-access
93
Woodcock, C. (2023, May 2). AI Is Tearing Wikipedia Apart. VICE.
https://www.vice.com/en/article/v7bdba/ai-is-tearing-wikipedia-apart
94
Senftleben, M. (2025, April 22). The TDM Opt-Out in the EU 3 Five Problems, One Solution. Kluwer
Copyright Blog.
https://copyrightblog.kluweriplaw.com/2025/04/22/the-tdm-opt-out-in-the-eu-five-problems-one-soluti
on/
23
