Copyright: AI, intelligence or theft?

Provocatively, one could question, at this point, the use of word ‘intelligence’ in the acronym ‘AI’. It is, on the face of it, a very sophisticated program designed to scrape data and repackage it for publication. That scraping presents a problem with respect to copyright law and how we view the legitimacy of original and derivative works.

This view has been taken 17 authors, including John Grisham, Jodi Picolut, David Baldacci and George RR Martin amongst others, in a recent Federal Court filing in New York. The suit alleges “flagrant and harmful infringements of plaintiff’s registered copyright” in a “systematic theft on a mass scale”. The CEO of the Authors Guild, Mary Rasenberger has added to this by stating: “To preserve our literature, authors must have the ability to control if and how their works are used by generative AI.”

In Federal Court cases in California involving comedian Sarah Silverman and author Paul Tremblay, OpenAI asked the Court to dismiss the proceeding because they “misconceive the scope of copyright, failing to take into account the limitations and exceptions (including fair use) that properly leave room for innovations like the large language models now at the forefront of artificial intelligence.”

What is topical from the claim made by OpenAI and others AI actors is whether the produced AI content can be considered ‘original’ for the purposes of copyright law and does the reproduced work fall within ‘fair use’ or legitimate derivative works. If the AI program is designed to scrape material from other copyright materials, the threshold question is whether the reproduced compilation is ‘original’ as required by the UK (Copyright, Designs and Patents Act 1988 [CDPA]), Switzerland (Federal Act on Copyright and Related Rights [CopA]), Australia (Copyright Act 1968 [AUS Act]) and other countries. The answer may fall to a question of degree in all the circumstances based on the original text that is alleged to have been infringed.

However, the question may not have to be entertained in Switzerland and Australia as copyright exists in the original works of human creation (Article 2(1) CopA and section 32 of the AUS Act). In the US, District Court Judge Beryl A. Howell recently found that the US Copyright Office was correct in denying copyright protections to literary content entirely produced without human involvement. The question then necessarily rests on how much human involvement is required to qualify an author of a claimed copyright work? How much involvement does the human need to exert over the program or process? We see this issue in the music industry where AI tools are often used for creative assistance. Arguably the human author is using creative expression to generate an original work. Again, another area of debate and further argument.

In Infopaq International A/S v Danske Dagblades Forening, Case C-5/08, ECLI:EU:C:209:465, the European Court of Justice stated that to benefit from copyright protection, the work must reflect “the author’s own intellectual creation”. This inputs a need to assess the actual input by a human into the claimed work. It is expected that subjective arguments will arise as to the creative input by a human into the instructions provided to the AI program to create the ‘original’ work. Interestingly, in the UK under the CDPA the law suggests that AI computer generated works can be protected by copyright.

What is clear is that this examination into what is a copyrightable work with respect to AI generated material is going to be vexed as we grapple with the onward march of the technology. Noting the existential issues in the letter from Max Tegmark of the Future of Life Institute from March 2023, “Should we develop nonhuman minds that might eventually outnumber, outsmart, obsolete and replace us? Should we risk loss of control of our civilisation?”. Whilst that question may not be directly relevant to the question of whether something is a copyrightable work, it will be something playing on the mind of judges and legislators as this issue unravels.