Who is the author — you or AI? Tensions rise over copyright protection
March 22, 2023 - The patent legal community continues to wait on further judicial and legislative action regarding whether artificial intelligence (AI) can be listed as an inventor on a patent application. The applications filed by computer scientist Stephen Thaler — involving a system that generated prototypes for a novel beverage holder and light beacon without input from him — brought attention to the topic of patent inventorship and AI, with Thaler urging patent offices to allow inventions conceived by the AI to be protected by patents. See, "U.S. scientist hits another dead end in patent case over AI 'inventor'," Reuters Legal News, Oct. 20, 2022.
However, U.S. and foreign jurisdictions disagreed with Thaler and continue to require an inventor to be human. Thaler has petitioned the Supreme Court to review the decision of the U.S. Court of Appeals for the Federal Circuit, although the outcome is not expected to change.
While the USPTO is open to the possibility of AI-assisted inventorship and has recently opened public comments on the topic (Request for Comments Regarding Artificial Intelligence and Inventorship, 88 Fed. Reg. 9492 (Feb. 14, 2023), legislative action will likely be needed to adapt the current patent laws to the quickly evolving world of AI.
In the meantime, the use of AI in generating artwork and software has been at the forefront of copyright related news. AI systems, such as ChatGPT, have received widespread (and controversial) attention due to both the process of training such AI systems using input of previously created artwork or software, as well as the work product generated by the AI. Although the benefits of ChatGPT and similar AI systems are clear, the underlying process of developing AI systems remains questionable.
From a software perspective, Microsoft, GitHub and OpenAI have each been named in a suit that alleges violations of copyright laws based on their reliance on open-source code as input for training of the AI. (Doe 1 et al. v. GitHub, Inc. et al, Docket No. 4:22-cv-06823 (N.D. Cal. Nov. 3, 2022) and Doe 3 et al v. GitHub, Inc. et al., Docket No. 4:22-cv-07074 (N.D. Cal. Nov. 10, 2022)).
In several instances, as alleged in the Microsoft, GitHub and OpenAI suit, open-source code may have been used to train the AI and the software generated by the AI potentially incorporated portions of the open-source code without acquiring a license or providing attribution to the original author(s).
In the artistic community, AI has been trained based on artwork created previously by others to generate "its own" artistic works. Again, the topic of the training process and the reliance of AI on copyrighted materials is controversial. As an example, Getty Images has sued Stability AI, alleging it has been trained based on copyrighted materials without obtaining a license. The complaint alleges that Stability AI has copied photographs from Getty Images' collection, along with the associated captions and metadata, and used this information to train its AI model. (Getty Images (US), Inc. v. Stability AI, Inc., Docket No. 1:23-cv-00135 (D. Del. Feb. 3, 2023)).
However, if an individual can walk through a gallery and use the photographs/paintings as inspiration for creation of a copyrightable work of art, is AI unable to do the same? The legal community awaits input from courts on these and similar questions, and whether the software or artwork relied upon for training is consistent with fair use.
One argument that repeats in these legal discussions is "incentive." If training of AI based on copyrighted/protected materials is permitted and the generated work product arguably includes elements of the copyrighted/protected materials, what is the incentive for software developers and artists to generate work if it will be freely usable by AI without providing attribution to the original author?
However, a similar argument can be presented for the incentives on the use of AI. If software/artwork generated by AI is not protectable via copyright because only a human can be an author under current copyright laws, and similarly inventions generated by AI are not protectable via patents, what is the incentive for the further development and integration of AI? Will such lack of author/inventor incentive stagnate AI innovation? See, e.g., United States Copyright Office, Correspondence ID 1-3ZPC6C3; SR # 1-7100387071, at 6 (Feb. 14, 2022) (concluding that USCO "policy and practice makes human authorship a prerequisite for copyright protection"); Thaler v. Vidal, 43 F.4th 1207, 1212 (Fed. Cir. 2022) ("[T]he Patent Act, when considered in its entirety, confirms that 'inventors' must be human beings.").
The legal community awaits answers and guidance on these topics. However, similar to the decisions involving AI and patents, the short-term outlook for those in favor of IP protection for AI is not optimistic. As an example, although the U.S. Copyright Office (USCO) initially allowed registration to Kristina Kashtanova of a comic book with images generated by AI, the USCO backtracked on this decision and clarified in February 2023 that the registration is for the comic book as a whole and for Kashtanova's text, but not for the individual images generated by AI.
This USCO decision and the pending lawsuits relating to AI and copyrights have pushed the USCO to review the current U.S. copyright laws relating to authorship. While the result of this USCO review is uncertain, it appears that copyright and patent laws may take the same route involving an extended waiting period until Congress takes legislative action to adapt to the inevitable development and integration of AI.