How Do Intellectual Property Rights Apply to Generative AI Outputs?

The following is an edited version of the content presented in the video.

Generative AI tools are made available under terms of use, which typically indicate who owns the outputs generated by the tool. In most cases, the terms indicate that users own the instances of outputs they prompt.

US courts have stated, however, that a computer cannot be the author of a copyright or an inventor on a patent. It follows that, if the creative or inventive aspect of an output was generated by the generative AI model, users cannot protect such outputs with a copyright or a patent — and therefore cannot invoke copyright or patent laws to prevent other parties from using generative AI outputs. Note, however, that inventive techniques of using a generative AI model are protectable by patent (assuming they meet the other criteria) and input prompts used to train or use a generative AI model are protectable by copyright.

Theoretically, the outputs of the generative AI tool could be deemed a trade secret. But again, this status may depend on how the tool’s terms of use treat inputs and outputs. For something to be deemed a trade secret, the owner must take appropriate actions to maintain its confidentiality, typically through physical, technical, or contractual means. In most cases, terms of use indicate that inputs and outputs will be fed back into the generative AI model to produce new inputs and outputs. As soon as that happens, trade secret protection could be lost. 

It is also important to remember that users can be held responsible for IP violations that result if they use outputs that were created based on protected IP, including software code.

Users should be cautious about using code that is produced by generative AI tools because it may be very difficult (if not impossible) to determine what source code the tool may have used to generate the code. If the user can’t identify the source code, the user can’t know the terms that govern that code’s use. The source code could be subject to an open source license, which could stipulate certain restrictions (such as allowing only non-commercial use) or obligations (such as requiring that the output code itself be made available under an open source license).

For more details, see “Three Steps to Manage Copyright Risk When Using Generative AI to Write Code.”