AI and Intellectual Property: Protecting What Your Company Creates

As companies build AI into their products and their day-to-day work, the question of who owns what AI helped create has become a business question as much as a legal one. Patents, copyrights, trademarks, and vendor contracts each treat AI-assisted work differently. These issues tend to surface at the worst possible moment, such as during an acquisition, a financing, or an effort to enforce rights against a competitor. Companies that deal with them early are generally in a much stronger position than those that find out about them during due diligence or litigation.

One important consideration is drafting patent claims around who will actually perform the claimed steps. AI agents increasingly act on behalf of users, so the steps of a method may be split among a customer’s agent, a third-party platform, and the company’s own servers. A method claim is generally infringed only when every step is performed by, or can be attributed to, a single party (sometimes referred to as divided infringement). A patent whose claims require several independent actors may be hard to enforce against anyone. Where possible, claims should be structured around what a single actor does, such as the company operating the server or providing the software.

Companies should also document the human contribution to inventions developed with AI tools. Only natural persons can be named as inventors, and the USPTO applies the same inventorship standard whether or not AI was used, asking who conceived the claimed invention. If inventorship is later questioned, records showing what the engineering team conceived, as opposed to what an AI tool suggested, can be important. The question may come from an accused infringer, a potential acquirer, or a proceeding before the Patent Office.

Another important consideration is that companies should not assume they own AI-generated content. Copyright protection requires a human author. In March 2026 the Supreme Court declined to review the D.C. Circuit’s decision in Thaler v. Perlmutter, which upheld that requirement. A logo, marketing image, or block of code generated by AI without meaningful human creative contribution may receive limited or no copyright protection. Human selection, arrangement, and modification of AI output can still be protected if sufficiently creative, so it is worth keeping track of what the team actually contributed to key assets.

Companies should also avoid using AI-generated mockups as evidence of trademark use. Before a product launches, it can be tempting to generate an image of the branded product and submit it to the USPTO as a specimen of use. The USPTO refuses mockups, renderings, and digitally created or altered specimens that do not show the mark in actual use in commerce. A registration that issues based on an improper specimen may be open to challenge years later, often just when the owner needs to enforce it. If a product has not yet launched, an application based on intent to use the mark is generally the better approach.

Finally, companies should review the terms of the AI tools their employees use rather than simply accepting them. Who owns the outputs, whether the vendor may retain company data or use it to train its models, and whether the vendor will indemnify the company against third-party infringement claims are business terms. They can often be negotiated, particularly in enterprise agreements. These terms are much easier to address before a tool is widely adopted than after company data and work product have already passed through it.

TAKEAWAY

AI can help a company build products and content faster, but it can also leave gaps in the intellectual property the company believes it owns. Companies using AI should draft patent claims so that every step is performed by, or attributable to, a single party, and should document the human contributions to their inventions and creative works. They should also file trademark applications on a basis that matches actual use, and negotiate AI vendor terms before employees adopt the tools. Dealing with these issues early costs far less than finding them during due diligence or enforcement.

Add a comment

Type the following characters: romeo, papa, tango, november, three

* Indicates a required field.

Subscribe

Recent Posts

Contributors

Archives

Jump to Page

Necessary Cookies

Necessary cookies enable core functionality such as security, network management, and accessibility. You may disable these by changing your browser settings, but this may affect how the website functions.

Analytical Cookies

Analytical cookies help us improve our website by collecting and reporting information on its usage. We access and process information from these cookies at an aggregate level.