At the time of this writing, generative artificial intelligence (AI) is taking the world by storm, and legal issues abound. Artists are suing AI art-generating companies for copyright infringement. Getty Images is suing for misuse of its images. Software developers are suing Microsoft and OpenAI for breach of contract, copyright infringement and other statutory claims. Technologists are testing patent and copyright laws in multiple countries to gain intellectual property (IP) protection for AI-generated work with no human author or inventor. On the other hand, artists are learning how much human contribution is necessary before a work is copyrightable under United States law.
These legal issues may temper some startups' enthusiasm for the use of AI. Legal disputes can arise at any point along the way, and many startups haven't planned for AI issues that can lead to claims and litigation. This article will examine emerging issues leading to disputes that startups and their legal counsel should be aware of, and why alternative dispute resolution (ADR) may be especially important in these unpredictable times.
Three Emerging Issues
There are three emerging issues in the AI space that may affect startups: lawsuits against generative AI services for copyright infringement, issues regarding IP with no human author or inventor, and the protection of AI-generated work with human modification. These may or may not affect how a startup uses generative AI.
Suing AI Services
Lawsuits against providers like GitHub and Stability AI allege that their AI-generative services misused copyrighted works to generate code and art. The GitHub complaint alleges that GitHub's product, Copilot, sometimes reproduces copyrighted code verbatim, without attribution or copyright notices. Similarly, Getty Images alleges that Stability AI has infringed on its trademarks by reproducing its watermark on some images that Stability AI generated.
It may be years before the law decides whether seeding AI-generated art and code with copyrightable works is fair use. The implications for a startup are clear: If AI can reproduce copyrighted works or company trademarks, a startup must have controls in place to avoid infringement. Although current litigation targets large entities like Microsoft, nothing prevents copyright holders from enforcing their rights against startups and small companies as well.
Protecting AI-Generated Works
Other litigation concerns whether AI-generated works are patentable or copyrightable. Both the U.S. Copyright Office (USCO) and U.S. Patent and Trademark Office (USPTO) have released guidance on AI-generated work. Both have concluded that a human being must author or invent the work to be protected. In Thaler v. Vidal, the Federal Circuit Court of Appeals has sided with the USPTO, finding that only human beings can seek patents as "inventors." The United States Supreme Court recently denied certiorari. Internationally, Thaler and the Artificial Inventor Project (AIP) are testing this issue in 18 jurisdictions. South Africa has granted such a patent. The U.K. Supreme Court just heard oral argument.
For many startups, the issues at stake in the AIP cases may be less profound. The AIP is specifically testing patent law to protect works without a human inventor. In the commercial world, many startups will take the practical step of listing a human inventor, even if AI generated much of the work.
That brings us to the last issue: How much human work is required before IP becomes copyrightable or patentable? There seems to be no current litigation on this issue, but that could change. Recently, an artist named Kristina Kashtanova applied for a copyright for her comic book, "Zarya of the Dawn." After granting a copyright, the USCO discovered that Kashtanova created the comic's artwork with Midjourney, an AI art generator. Based on this new information, the USCO concluded that Kashtanova was the author of the text and layout of the comic book, but not the images. Consequently, the USCO denied Kashtanova's copyright as to the images, finding that Kashtanova had not modified the AI artwork enough to make it her own.
The line between generated IP that cannot (currently) be copyrighted or patented and human-modified IP that merits U.S. copyright or patent protection has yet to be drawn. But startups should be aware and have mechanisms in place to ensure that there is enough human work to argue for protection.
Founder agreements should address the use of generative AI by the founders themselves in creating company IP, as well as its effect on ownership and enforcement of intellectual property rights. Founders should not overestimate the impact of Thaler on their works. Thaler is testing the law, asserting that an AI-generated work without a human inventor deserves patent protection. Based on that decision, it's clear that founders may need to build protocols to ensure human authorship.
Kashtanova's situation raises greater concerns. If the USCO denied a copyright for insufficient human contribution, founders must develop protocols to require and document the human contribution to their IP. Founder agreements may require written approval by some or all founders for AI-assisted development of products and services. Founders may want to consider disclosure of which generative AI tools they use and develop written protocols to document the human contribution for invention and creation of company IP.
Startups should also address the emerging issue of use of AI by employees. Current employment contracts probably protect the trade secret aspects of AI-generated work. Nevertheless, in this age of AI uncertainty, startups should regularly consult legal counsel to determine who owns (or co-owns) intellectual property and related rights. Although an AI-generated secret is probably still a trade secret, this assumption has yet to be tested. Startups will want to know who, if anyone, owns the IP and if an employee could use that IP post-employment.
Moreover, employee conduct may bring a startup face to face with the uncertainties of AI-related infringement litigation. Startups should have robust protocols to minimize the chance unintentional or inadvertent infringement. Additionally, inadequately trained employees may neglect to document their human contributions, which might prevent copyright or patent protection for company works.
Benefits of ADR
The pace of development for AI-generated work is breathtaking. The law, as usual with tech matters, plays the tortoise to tech's hare. These uncertain times highlight the benefits of ADR for startups. While the law is uncertain, startups may benefit from using hand-picked neutrals with relevant skills to mediate or arbitrate their disputes. Moreover, private mediation and arbitration can mitigate the development of bad case law. Lastly, the confidential nature of ADR can help protect a company's reputation and the privacy of its trade secrets.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.