(March 31, 2023) - Robert A. McFarlane of Hanson Bridgett LLP explains that trade secret principles offer an attractive alternative to patent and copyright for AI-related subject matter, in light of recent rulings requiring human inventors when seeking patents and copyrights.

Artificial Intelligence suddenly seems to be everywhere. ChatGPT is writing human-sounding sermons, news updates, and answers to law school exam questions, while Dall·E is generating images ranging from the lifelike to the surreal in response to virtually any prompt.

With much less fanfare, AI has already become ubiquitous in myriad ways. AI curates social media feeds and generates purchasing suggestions to fill internet shopping carts. AI saves lives by identifying potential pharmaceutical compounds and by quickly and accurately interpreting medical scans and images. And AI is learning to drive.

AI is even making inroads to the tradition-bound and technology-resistant legal profession. Lawyers are using AI to streamline eDiscovery reviews and, more experimentally for now, to create first drafts of common legal documents.

Indeed, AI feels much like the internet did in the late 1990s. Its time has arrived. It is being widely adopted. And it is transforming everything it touches in ways that are impossible to predict.

Like so many tech companies before them, innovators working with AI are seeking to protect the valuable intellectual property at the heart of their business models. However, the current IP landscape complicates efforts to protect AI-related subject matter.

One of the biggest obstacles to protecting material generated by AI can be the lack of a human creator. This issue has been explored by an AI developer named Stephen Thaler through his efforts to patent and copyright creations generated by an AI system called DABUS.

Thaler filed patent applications seeking patent protection for two inventions — a food container incorporating fractal geometry and an emergency beacon that pulsed at a frequency determined by fractal dimensionality — that the DABUS AI purportedly created without human contribution.

The USPTO rejected the applications as incomplete for the simple reason that they lacked a human who could be named as the inventor. When Thaler appealed to the U.S. Court of Appeals for the Federal Circuit, the court chose to avoid weighty "metaphysical matters" regarding "the nature of invention or the rights, if any, of AI systems." Thaler v. Vidal, 43 F.4th 1207, 1210 (Fed. Cir. 2022).

Instead, the court treated the issue as one of straightforward statutory interpretation and ruled that the patent laws clearly limit inventorship to human beings. Id. Without a human inventor, DABUS' inventions were left as unpatentable orphans beyond the protection of the patent system.

Thaler has filed a writ of certiorari, but given the unambiguous statutory language and decades of case law limiting inventorship to natural persons, it seems unlikely that the Supreme Court will intervene. "U.S. Supreme Court asked to decide if AI can be a patent 'inventor,'" Reuters Legal News, March 17, 2023, https://reut.rs/40vfoao.

Thaler's efforts also exposed difficulties in protecting AI-generated output through copyright registrations. Thaler asked the United States Copyright Office to recognize DABUS as the author of a two-dimensional work of art entitled "A Recent Entrance to Paradise," a fanciful image of a railroad track disappearing into a peaceful floral setting. Once again, Thaler claimed that DABUS produced the subject matter without any creative input from a human actor. See https://bit.ly/3lsAwzb.

Citing long-standing precedent, the USCO ruled that copyright protection was limited to works that are the product of human authorship and denied the requested registration. Id. at 3. Thaler is challenging this ruling in the federal district court case Thaler v. Perlmutter, case number 1:22-cv-01564 (U.S. District Court for the District of Columbia).

Because Thaler asserted that the subject matter in both instances was created without human contribution, the decisions involving DABUS did not determine whether subject matter created partly by an AI or by a human using AI as a tool could be protected. The USCO took up these issues when Kristina Kashtanova sought a copyright registration for a comic book that she created using the Midjourney artificial intelligence. See https://bit.ly/3TF9uBd. The work at issue included text written by Kashtanova and images created by Midjourney that Kashtanova had selected, coordinated and arranged into the final compilation.

The USCO determined that Kashtanova could register a copyright for the work's text and for the "selection, coordination, and arrangement of text created by the author and artwork generated by artificial intelligence." Id. at 12. The text was the product of human authorship and both aspects of the issued registration reflected sufficient creativity to be protected by copyright. Id. at 4-5. However, the registration explicitly excludes "artwork generated by artificial intelligence." Id.

Just as in Thaler, the AI-generated images lacked a human author. Moreover, Midjourney's output could not be meaningfully predicted by its users, which distinguished it from other tools used by artists in creating works that can be protected by copyright. Id. at 10.

The subject matter created by DABUS and Midjourney demonstrate limits on protecting AI-generated content through patents and copyrights. Unfortunately for AI developers, the Supreme Court's decision in Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014), also complicates gaining patent protection on the AI systems themselves.

Alice created a two-step analysis to determine whether a claimed invention is eligible for patent protection. Under the first step, the court asks whether a patent claim is directed to ineligible subject matter such as an abstract idea, law of nature, or natural phenomenon. If the answer to that question is "yes," the court must then ask whether the claimed invention adds an "inventive concept" sufficient to transform the ineligible subject matter into a patent-eligible application.

The USPTO and the courts have found a multitude of patents drawn to software-based inventions and inventions that rely on the use of algorithms are patent-ineligible. Since AI systems are broadly based on software incorporating algorithms, inventors seeking to patent AI-related advances must carefully consider Alice in drafting their patent claims and in deciding whether to seek patent protection at all.

Given the limitations on patenting and copyrighting AI-related subject matter, trade secret principles offer an attractive alternative. Trade secret law protects "all forms and types of financial, business, scientific, technical, economic, or engineering information" so long as the information's owner has "taken reasonable measures to keep such information secret," and "the information derives independent economic value ... from not being generally known to, and not being readily ascertainable through proper means by, another person who can obtain economic value from the disclosure or use of the information." 18 U.S.C. §1839(3).

Requirements centering on secrecy preclude trade secret protection for non-confidential outputs of AIs such as ChatGPT or Dall·E. However, trade secret law is well adapted to protect a host of AI-related material including training data, AI software code, input parameters, and AI-generated output that is intended only for internal and confidential use. And, significantly, there is no requirement that a trade secret be created by a human being, and AI-generated material is treated like any other information. See. e.g., 18 U.S.C. §1839(4) (defining trade secret owner).

The current legal landscape presents a complex environment that demands a pragmatic and nuanced approach to protecting AI-related intellectual property. Copyright can protect AI software code, human compilations and arrangements of AI-generated images. Patents can be used to protect aspects of AI systems that can pass the two-step Alice test.

Trade secrets can be used to protect confidential features of AI systems and their outputs, so long as reasonable steps to maintain their secrecy are taken, the material remains confidential, and competitors are not able to derive the material through legitimate or independent means. Thus, companies using or generating valuable AI will need to determine the most valuable aspects of their systems and their outputs and draw from a range of IP theories to tailor their intellectual property strategies and ensure protection for their most important AI-related developments.