Generative Artificial Intelligence (AI) tools have been a hot topic of both news and conversation since OpenAI launched ChatGPT in March 2023. In addition to the possibilities these tools present, there has been much discussion about their limitations and risks. Let’s take a look at some things you should be aware of from a copyright and citation perspective.
The big question is who owns the copyright for AI-generated materials?
As yet, there are no clear answers.
OpenAI, for example, claims to own the copyright for content produced by ChatGPT and its image generator, Dall-E. OpenAI assigns its users the right to use the content for any purpose, including commercial use (OpenAI, 2023a), but do they really have the authority to do so?
ChatGPT and other generative AI tools are ‘trained’ on huge amounts of data “scraped from the internet” (CBA National, 2022). Although this content is freely accessible, much of it is likely copyright protected, and that’s where things get murky. According to the Canadian Bar Association, clarity will likely have to be provided by court decisions—in Canada and worldwide—to address questions such as:
(CBA National, 2022)
There are already several high profile court cases involving generative AI. For example, Getty Images is suing Stability AI for allegedly using its content without permission to ‘train’ the image generator, Stable Diffusion (Brittain, 2023).
OpenAI (2023b) acknowledges that ChatGPT “sometimes writes plausible-sounding but incorrect or nonsensical answers.” This is sometimes referred to as hallucination and fake citations are one example.
University of Waterloo Library and Memorial Sloan Kettering Cancer Center Library, among others, have tested the reliability of citations produced by ChatGPT. I thought I’d try it myself.
I asked ChatGPT to write a paragraph on the use of carbamazepine in benzodiazepine withdrawal and to include a reference list.
This was the response:
Carbamazepine is a medication commonly used in the management of benzodiazepine withdrawal, particularly in cases of long-term benzodiazepine dependence. It belongs to the class of anticonvulsant drugs and acts by stabilizing neuronal membranes and reducing excessive neuronal firing. Carbamazepine helps alleviate withdrawal symptoms such as anxiety, insomnia, restlessness, and irritability, making it a valuable adjunct in the detoxification process. The drug's mechanism of action involves modulating the gamma-aminobutyric acid (GABA) neurotransmission system, which is disrupted during benzodiazepine withdrawal. While carbamazepine has demonstrated efficacy in managing benzodiazepine withdrawal symptoms, it should be used under close medical supervision due to its potential for side effects and interactions with other medications. Further research is needed to optimize dosing and duration of carbamazepine therapy in benzodiazepine withdrawal.
Reference List:
I am not a subject matter expert, so I won’t speak to the accuracy of the paragraph (though I would welcome input from readers). As for the citations, at first glance they looked promising—formatted correctly, reputable journal titles—but looks, as the saying goes, can be deceiving.
Of the four citations, only one is even partially correct. The others are complete fabrication!
A search of PubMed® and Google Scholar indicates that references 1, 3, and 4 do not exist. The DOIs provided by ChatGPT are for different and unrelated articles. Reference 2 IS a real article from the British Journal of General Practice, but the authors and the DOI are incorrect.
If you use generative AI tools for any purpose:
If you have questions about copyright:
Further Reading:
Alkaissi H, McFarlane S I (February 19, 2023) Artificial Hallucinations in ChatGPT: Implications in Scientific Writing. Cureus 15(2): e35179. doi:10.7759/cureus.35179
Flanagin, S., Bibbins-Domingo, K, Berkwits, M. & Christiansen, M.A.. (2023, Jan. 31). Nonhuman “authors” and implications for the integrity of scientific publication and medical knowledge. JAMA 329(8), 637-639. doi:10.1001/jama.2023.1344
References (not AI generated)
1. Brittain, B. (2023, Feb. 6). Getty Images lawsuit says Stability AI misused phots to train AI. Reuters. https://www.reuters.com/legal/getty-images-lawsuit-says-stability-ai-misused-photos-train-ai-2023-02-06/
2. CBA National/ABC National. (2022, Dec 7). The legal status of generative AI. The Canadian Bar Association. https://nationalmagazine.ca/en-ca/articles/law/hot-topics-in-law/2022/the-legal-status-of-generative-ai
3. CCH Canadian Ltd. v. Law Society of Upper Canada. [2004] 1 S.C.R. 339, 2004 SCC 13. https://scc-csc.lexum.com/scc-csc/scc-csc/en/item/2125/index.do
4. O’Brien, M. (2023, Jan. 6). What is ChatGPT and why are schools blocking it? The Globe and Mail. https://www.theglobeandmail.com/world/article-explainer-what-is-chatgpt-and-why-are-schools-blocking-it-2/
5. OpenAI. (2023a). Terms of use. https://openai.com/policies/terms-of-use
6. OpenAI. (2023b). IntroducingChatGPT. https://openai.com/blog/chatgpt
Librarian Educator
Central Zone
0 Comments.