Artificial Intelligence in the Modern Workplace: Safeguarding Source Code Generated with AI Assistance

Generally, an employer owns all rights in software code created by its employee in the scope of their employment. As outlined in the last edition of this series, this general rule typically applies to independent contractors so long as the term is included in a written contract. Even where an employee uses generative artificial intelligence (“GenAI”) to either generate an entire set of code or review their code for errors, the employer typically maintains ownership of the code. This edition evaluates the manners in which employers can protect the code they own.

Standard Protections for Source Code

As has long been the standard, there are various ways a business can employ standard intellectual property protections (copyright, patent, or trade secret) to protect their proprietary source code. Copyrights can be used to protect the literal components of the source code. However, it does not protect the functional aspects of the software, and thus cannot protect against a third party’s use of the independently developed algorithms or the functional equivalent of the source code. A copyright infringement claim is based on the actual similarities between the source code and the allegedly infringing source code.

Patents, on the other hand, protect the functional aspect of source code. This means that a patent can provide protection against the use of the proprietary technology, but unlike copyright protection, the literal component of the code is not directly protected. Copyright protection begins at the moment the code is fixed in a tangible medium of expression, while patent protection must be obtained from the U.S. Patent and Trademark Office (usually taking 2-3 years to obtain patent protection).

Trade secrets can often be an effective way to protect a company’s proprietary information and source code. Where reasonable steps are taken to keep the source code secret, this method of protecting the intellectual property can provide appropriate protection. One benefit – as compared with filing a copyright or seeking patent protection – is that the company does not need to file the information with a governmental agency. The risk, however, is that if the trade secret is publicly disclosed or reverse engineered, then one could lose the trade secret protection. 

These three methods of protection are all based on the assumption that the source code was created entirely by a human. While an employer will likely maintain ownership in code created with GenAI assistance, the protectability of that code may not fit well in the current standard intellectual property protection.

Protectability of AI Generated Code

Copyrightability

The courts are beginning to address the question of who owns works generated with AI assistance. The U.S. District Court for the District of Columbia found that human prompting of AI-generated works does not satisfy the “authorship” requirement for copyright protection. Thaler v. Perlmutter, No. CV 22-1564 (BAH), 2023 WL 5333236 (D.D.C. Aug. 18, 2023). Because the Copyright Act of 1976 requires human authorship, the court found that a piece of art generated by AI was not eligible for copyright protection, regardless of the fact that a human prompted the AI. While not specifically addressing ownership concerns over the creation of code, the same principles apply. It would appear that the courts are hesitant to find copyright protection in works generated by AI.

This decision, however, does not act as a complete ban on any and all AI-generated work. The district court was instead more nuanced in its decision, finding that “the Copyright Office acted properly in denying copyright registration for a work created absent any human involvement.” This leaves open the question of just how much human involvement is necessary for a work to qualify for copyright protection. While an employer may own any code created by an employee, even where GenAI was used, the employer’s ability to protect such code through copyright is in flux. In order to mitigate the risks associated with using GenAI in code creation, businesses should consider implementing policies that manage how GenAI can be used in assisting in creating such code.

Patentability

The U.S. Court of Appeals for the Federal Circuit has also rejected a scientist’s applications for patents on inventions created by an autonomous system. The scientist, Stephen Thaler, argued that his autonomous system – Device for the Autonomous Bootstrapping of Unified Sentience – is “natural and sentient” and thus eligible for inventorship. The Federal Circuit rejected this argument and stated that the Patent Act requires an “inventor” to be a natural person. Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), cert. denied, 143 S. Ct. 1783, 215 L. Ed. 2d 671 (2023). Again, while an employer would likely own any rights in code created with GenAI assistance, patent protection may not be an available means for protecting the code.

Trade Secret Protection

The remaining form of possible protection is trade secret protection. Where GenAI is used to assist in the generation of proprietary code, and such code provides an economic advantage to the business based on the fact that it is unknown to the general public, trade secret protection is an option. This is where company policies are important. Because a trade secret’s value is connected to the secrecy of the information, it is important that appropriate workplace policies and practices are implemented to safeguard this information and keep it secret.

It is important that a business properly evaluates and explores these avenues of protecting its proprietary code. As the courts continue to address the issues of protectability of content created using GenAI, it is important that a businesses’ strategy for protecting its proprietary code evolves with changes in the legal landscape.