Are there rules around claiming copyright if I write a novel or a research paper using generative AI? If I ask a large language model like ChatGPT or Bing Chat to put together a paragraph or two on some topic and include it in my writing, can I claim it as all my own work?
Governments, authorities and lawmakers around the world are wrestling with how artificial intelligence can be corralled into being useful, while limiting the harm it could cause.
The question that remains to be answered is who can claim to be the author of a work that has incorporated material produced by generative AI.
How is output from generative AI treated under copyright laws?
Under current Australian intellectual property law, a literary work must originate through an author’s independent intellectual effort.
But the generative AI output from a large language model such as ChatGPT is the result of trawling through masses of other authors’ writings found on the internet to produce answers to questions put by the user.
A ChatGPT user may ask for a vivid description of a storm at sea and the answer could include phrases that were compiled from authors such as Hemingway, Monsarrat and Shakespeare.
The Commonwealth Copyright Act 1968 states that a work must originate with a human author. ChatGPT is not a human, so it does not qualify for authorship.
A work that is predominantly created by a human with a bit of contribution from AI would probably be protected by copyright.
But if generative AI has done most of the work, with only a bit of tinkering by a human, then it most likely will not be subject to copyright.
How copyright currently works in Australia
The law says copyright provides legal protection for people who express original ideas and information in forms such as writing, visual images, music and moving images. (Please see Copyright basics, Attorney-General’s Department.)
Copyright is free and automatic under the Copyright Act. You don’t need to register for copyright in Australia. The moment an idea or creative concept is documented on paper or electronically, it is automatically protected by copyright.
Copyright does not protect ideas or information, but only the original expression of the ideas or information. There are “fair dealing” exceptions which enable some use of copyright material in work such as research, academic study and news reporting.
Copyright lasts for 70 years after the death of the author. After that it falls into the public domain and can be freely used by others.
Importantly where generative AI is concerned, copyright does not prevent someone else from independently producing the same work, as long as they came to it without copying from a copyrighted work.
Two authors sue OpenAI for breach of copyright
The Copyright Act does not deal with the concepts underpinning generative AI, such as data mining and machine learning, and it has yet to be tested properly in court.
In July 2023, two authors in the US sued OpenAI, the company which owns ChatGPT, claiming it had breached their copyright when it used their books to “train” ChatGPT without their permission. (Please see Authors sue OpenAI, allege their books were used to train ChatGPT without their consent, CNBC, 5 July 2023.)
However, it will be a challenge for the authors to prove exactly how OpenAI used their books, and whether they suffered financial loss. OpenAI has claimed it did not reproduce any of the copyrighted material in the authors’ books.
Further reading on artificial intelligence and generative AI
Inventiveness of ChatGPT poses risk of defamation
AI-generated deepfake images create bullying danger
Predicting recidivism – the questionable role of algorithms in the criminal justice system
Algorithms, artificial intelligence, automated systems and the law
Guilty or not guilty – could computers replace judges in a court of law?