Tread a Careful Legal Path Around AI Chatbots

This is the first in a periodic series of pieces examining some of the issues that are arising around the use of AI-generated content.

Copyright and Indemnification

A company, which will remain nameless, recently contacted GlobeSt.com about a commercial real estate service that incorporated ChatGPT’s natural language processing. An interesting concept, but when asked about potential issues of copyright, the firm didn’t respond.

This shouldn’t be surprising because no one would wish to be on the wrong side of a copyright lawsuit. The issue isn’t about one given company, or even a question about copyright, but a collection of conditions that together make serious due diligence and risk management necessary for companies, including those in CRE, that want to make effective commercial use of the large language model forms of artificial intelligence that seem to be appearing everywhere.

At the basic level, copyright is legal protection for covered works—writing, artwork, photos, architectural drawings, videos, and so on. Not for an underlying idea, but its specific expression. You generally can’t use copyrighted work without the permission of the rights owner, although there are exceptions known as fair use. News reporting, criticism, commentary, teaching, scholarship, and research can all be done to some degree. Commercial use doesn’t fall under the exemption.

The large language models are machine learning systems trained on available content. Significant bodies of work including much from large publishers are used for the statistical analysis that determines, within a given context, what words might more reliably come after one another. So far, this hasn’t been a legally challenged copyright issue because the researchers could point to fair use. In a commercial context, which include services being offered to businesses to speed content creation, this may no longer be the case. Lawyers from large publishers like Dow Jones have said that they are monitoring this closely. Given that AI use of images as a starting point are already in the courts, it isn’t a long stretch to imagine the same eventually happening regarding the language models.

What complicates all this is that ChatGPT in particular and likely similar systems include indemnification for any legal trouble as part of the terms of conditions of use. The financial implications of any legal challenges could bounce back onto the users. Whether that would hold up or not in a court is impossible to say, but anything you have to call in the lawyers, things start getting messy and expensive.

On March 16, the U.S. Copyright Office announced “a new initiative to examine the copyright law and policy issues raised by artificial intelligence (AI), including the scope of copyright in works generated using AI tools and the use of copyrighted materials in AI training.”

In its current take on the topic, the Copyright Office published the following in the Federal Register:

“In the Office’s view, it is well-established that copyright can protect only material that is the product of human creativity. Most fundamentally, the term ‘author,’ which is used in both the Constitution and the Copyright Act, excludes non-humans. The Office’s registration policies and regulations reflect statutory and judicial guidance on this issue.”

There can be room for the use of AI, like any other technology, in creating a copyrightable work, but it’s on a “case-by-case inquiry” of “whether the AI contributions are the result of ‘mechanical reproduction’ or instead of an author’s ‘own original mental conception, to which [the author] gave visible form.’”

In other words, if you use these systems to generate content, it’s not clear whether you can get copyright protection for them. If not, competitors would be able to freely use the same material.