The intersection of Artificial intelligence and Intellectual Property is complex. It involves several IP rights, some of which overlap in some cases: copyright, trademarks, patents, trade secrets/confidential information, and the right of publicity (and similar rights with different names). The situation has increased in complexity now that not only the input but also the output of Large Language Models (or LLMs) has allowed AI machines to produce potentially patentable inventions and content that looks like literary and artistic material that, in certain cases at least, a human author could have created. The following is a checklist for scholars and practitioners who are looking at an AI-related IP issue. Its main purpose is to help structure the discussion.
1. Is scraping/text and data mining legal?
a. Fair use in US (Google Books but reuse pattern different here. Impact of Warhol, esp for GenAI)
b. Fair dealing
c. EU (CDSM arts 3-4; obligation concerning sufficiently detailed summary in June 2023 draft of AI Act)
d. Japan (Art. 47 septies)
e. Singapore (computational data analysis; user must not “use the copy for any other purpose”)
f. Switzerland (scientific research)
g. Potential role of three-step test (Berne 9(2)/TRIPS 13)
2. Does the machine infringe when it produces a new “work”?
a. For reproduction: traditional analysis
b. For the right to prepare a derivative work in US, linked to issue 3, see paper #1 and Getty Images lawsuit
3. Can the machine be an author?
a. USCO and Federal Court decisions in US
b. Separate cases of “collaboration” from “pure” machine productions: can use notion of cause. See paper #2.
c. On normative issues, see paper #3.
4. Is the algorithm copyrightable?
a. Computer in source and object code protected (if written by humans). TRIPS 10.1; 17 USC §101, 106, 117
b. Algorithm subject to 102(b)/exclusion of ideas
5. Is data set copyrightable?
a. Issue of authorship (previous)
b. Most likely not original in any event
6. Rights Management Information
a. 17 U.S.C. 1202 (possible statutory damages–even for unregistered works; but P must show D knew, or, with respect to civil remedies under section 1203, had reasonable grounds to know, that removal/distribution would “induce, enable, facilitate, or conceal an infringement of any right under this title”
b. WIPO Copyright Treaty
1. Can a machine be an “inventor”?
a. No, say a vast majority of Patent Offices and courts. See paper #4 for cites and discussion.
2. Impacts on innovation?
a. See papers #2 and 4.
C. Trade secrets
1. Can datasets and algorithms be protected as trade secrets?
a. Can probably qualify as “information” if meets other requirements under TRIPS 39, UTSA etc.
b. No protection against reverse engineering or independent production
c. Possible clash with transparency and “explainability” obligations (eg People v. Wakefield, 107 N.Y.S.3d 487 (N.Y. App. Div. 3d Dept. 2019)
D. Right of publicity/INL
a. Protection of fame monetization vs privacy
b. Right to image can be fairly strict (eg Quebec)
c. Doesn’t protect style
E. Private international law/Conflicts
a. Jurisdiction (eg scraping vs access vs new production)
b. Possible antisuit injunction?
- AI Derivatives: The Application of the Derivative Work Right to Literary and Artistic Productions of AI Machines, 52:4 Seton Hall L Rev 1111 (2022)
- The Human Cause, in Research Handbook on Intellectual Property and Artificial Intelligence (R. Abbott, ed), (Edward Edgar, 2022) pp 21-38.
- The Machine as Author, 105 Iowa L. Rev. 2053-2106 (2020)
- Artificial Inventors.in Intellectual Property – A Global Project (S Frankel, M Chon, G B Dinwoodie, B Lauriat and J Schovsbo, eds) Edward Elgar 2023, Available at SSRN: https://ssrn.com/abstract_id=4410992
- Słok-Wódkowska, M. and Mazur, J. (2022) ‘Secrecy by Default: How Regional Trade Agreements Reshape Protection of Source Code’, Journal of International Economic Law, 25(1), pp. 91–109. doi: 10.1093/jiel/jgac005.