Usually, FERPA and Copyright issues are first of mind with me in my classroom, but for some reason I hadn’t really engaged my brain around these issues when it came to NotebookLM (Ah, young love! So intoxicating!!). It took an email from my friend, Saralyn, for me to shake myself out of my infatuation with this tool and get real about NotebookLM. Here’s what she wrote:
I‘ve been reading about Notebook LM and thinking about how to use it in my classes, specifically for research assignments. However, two of our librarians have expressed concern to me about possible legal violations when uploading copyrighted sources to LM. I don’t want to put my students or myself at risk. Do you talk with your students about how not to violate copyright?
Hmm. It appears our librarians, who are usually supernaturally correct in everything copyright, are somewhat incorrect when it comes to NotebookLM. I am assuming those stalwart guardians of grimoire, gazettes and granthamala had NotebookLM confused with lesser tools of the AI genre. It is correct to be concerned about uploading copyrighted and privileged information into a general AI chatbot like ChatGPT, Gemini, or Claude—as that material would then be included in the training data of a general LLM, but NotebookLM is different.
Copyright and NotebookLM
Unlike most general AI tools, Google’s NotebookLM is a slightly more pedantic Large Language Model (LLM) that utilizes RAG: Retrieval-Augmented Generation. RAG is an AI framework where the model retrieves information from a specific knowledge base (documents, databases, images, etc.) and then generates information based upon that material. This is why NotebookLM is different than most AI systems, and why it is so useful to me, as an educator.
For example, if I have a student upload my syllabus and ask questions to NotebookLM about my syllabus, NotebookLM is less likely to give the student hallucinated information, and it will also reference the specific passage in my syllabus where the information has been retrieved, so the student can directly access it. I don’t have to worry about the model accessing someone else’s syllabus and giving my student fraudulent information.
The second part of this advantage is that NotebookLM can access that information without having to drag it out into their LLM. This is why Google can publish a page that stresses that sources you upload stay private unless you choose to share a notebook, and that NotebookLM does not train on uploaded data (Workspace).
According to a public discussion held on the Association of College Research Libraries, uploading research into NotebookLM for private educational use is generally permissible under fair use principles if you have lawful access to the materials (Google NotebookLM—Copyright Questions). There are three important caveats to this idea, however:
You must only upload sources into NotebookLM that you have the right to use. In order to claim fair use, you can’t upload something from a “pirate” library or a source that specifically prohibits its use. This was established in Bartz v. Anthropic, where Judge William Alsup ruled that “while using copyrighted books to train artificial intelligence constitutes fair use due to its transformative nature, maintaining a permanent library of pirated content does not qualify for such protection” (Bartz et. al).
You must not publicly share the contents of a notebook. You are welcome to remix and reassemble that information, or quote from it, but you cannot share it in its entirety. In Author’s Guild v. HathiTrust, the 2nd Circuit Court of Appeals held that creating a database of source material for search and accessibility was transformative fair use because the information was ABOUT the works, not the consumption of their protected content (Cox). In addition, NotebookLM’s policy explicitly requires users to respect copyright laws and prohibits sharing copyrighted content without rights, with account termination for repeated infringement (Google Workspace Updates).
The information in the Notebook cannot be used in lieu of a resource currently marketed for that use. You don’t want to, for example, upload a textbook that you could have required students to use, then make videos and podcasts from that textbook for your students so they don’t have to purchase that textbook. That’s not only unfair to the author and the publisher, it is also unethical and would jeopardize educational fair use. A February 2025 ruling in Thomson Reuters v ROSS Intelligence in the 3rd Circuit Court of Appeals found that ROSS Intelligence used the Copyrighted headnotes created by Thomson Reuters as AI training data to create a legal research tool that would replace Thomson Reuters tool, and ruled against ROSS Intelligence in the case (Lawnext).
You are allowed, under Fair Use to keep a database of copyrighted materials for research and for transformative use, just as long as you obtained that material legally and you do not expose the protected content of those copyrighted materials to the general public (Authors Guild v. HathiTrust). So, what you are doing in NotebookLM is protected fair use.
NotebookLM is powerful, and it operates within a legal and ethical minefield where most AI tools fail. But, in order to stay within the law—instead of uploading that material into your notebook and then sharing it with your class, consider building your students’ AI skills AND staying within copyright constraints by discussing fair use with your students and then having them upload that material into their own private notebooks in order to gain insight through their own transformation of those materials into interactive learning opportunities.
When Publisher Terms Trump Fair Use
Even if your use feels like “Fair Use,” a publisher’s contract might say otherwise. This is the world of Text and Data Mining (TDM) clauses. Many publishers now include specific language in their Terms of Service that prohibits inputting their content into AI tools.
The landscape is inconsistent. For instance, academic journals like Sage and Liebert generally allow TDM for non-commercial research if you have lawful access. Others, like Wiley or Elsevier, have more technical stipulations. As librarian Amy Bergeron has noted, this creates a “scruples vs. funds” dilemma (Google NotebookLM). Research labs with significant funding can buy high-tier, private tools that negotiate these rights, while those without funds may be tempted to ignore TDM policies just to keep pace. As digital literacy advocates, we must be careful that our push for efficiency doesn’t lead us to ignore the contractual “fine print” that protects our library’s access. The good news is that you can upload those dense usage rights documents to NotebookLM and query them in order to understand the fine print.
The Output is Not Your Property
NotebookLM rocks the synthesis of complex material for teaching, learning, and research. Synthesis—the act of turning a Rumpelstiltskinian pile of research sources into the golden thread of insight—is the most time-consuming part of research, and the one thing that NotebookLM is especially good at managing and transforming into podcasts, explainer videos, infographics, slide decks, and data tables.
Here’s an example of how to use those tools as an educator to help your students make their NotebookLM use even more incredible. Here are some great NotebookLM prompts from educator, Tricia Friedman.
So, who owns that brilliant AI-generated study guide? According to the U.S. Copyright Office, not you. Current policy emphasizes the “centrality of human creativity.” The Office has made it clear that the “mere provision of prompts” is not enough to secure a copyright for AI output.
To claim ownership of a generated lesson plan, you must provide what the law calls “sufficient expressive elements.” For example, if you take a raw study guide from NotebookLM and substantively arrange, annotate, and integrate it into a larger, original course module, you have added the necessary “creative spark.” The AI can help you connect the dots, but the legal protection only applies to the way you, the human, choose to arrange them.
NotebookLM and FERPA
NotebookLM is classified as a “Core Service” for Google Workspace and Workspace for Education Accounts (NotebookLM Help). This means it comes with enterprise-grade protections. That means that if your educational institution has Google Workspace, you are golden! If you are using NotebookLM in a consumer (private) account, that data is not used to train models unless you EXPLICITLY provide feedback and allow that use. According to Google’s Privacy Hub, data in NotebookLM is not human-reviewed and—most importantly—is not used to train Google’s AI models without your explicit permission. According to Google “NotebookLM functions as a ‘closed system,’ meaning it grounds its responses only in the documents you upload and does not access the public internet for answers. This design prevents the unintentional exposure of sensitive information” (Lim).
However, there is a massive trap here for the unwary. These protections only apply if you are logged into your institutional account. If you or your students use a personal (@gmail.com) account to process student data or sensitive research, you are stepping outside that “closed-loop” safety net. In fact, Google’s own documentation notes that “public sharing” features are only enabled for consumer accounts, creating a significant privacy risk.
Using a personal account for institutional business isn’t just bad practice; it’s a potential FERPA violation waiting to happen. This is why, if you don’t currently have a Google Workspace for Education account at your school, you need to be very careful about adding sensitive, FERPA protected, or proprietary research information into NotebookLM—just to be extra careful. My institution does not currently have an Workspace Education account, so I have this conversation with my students.
Works Cited
Ambrogi, Bob. “Breaking: Federal Judge Rules Legal Research Startup ROSS Infringed Westlaw’s Copyrights, Rejecting Fair Use Defense | LawSites.” Commercial. Lawnext.Com, 11 Feb. 2025, https://www.lawnext.com/2025/02/breaking-federal-judge-rules-legal-research-startup-ross-infringed-westlaws-copyrights-rejecting-fair-use-defense.html.
“Authors Guild, Inc. v. HathiTrust , No. 12-4547 (2d Cir. 2014).” Justia Law, https://law.justia.com/cases/federal/appellate-courts/ca2/12-4547/12-4547-2014-06-10.html. Accessed 5 Mar. 2026.
Bartz, Andrea, et al. UNITED STATES DISTRICT COURT NORTHERN DISTRICT OF CALIFORNIA.
Cox, Krista. “Second Circuit Affirms Fair Use in Google Books Case.” Association of Research Libraries, 16 Oct. 2015, https://www.arl.org/blog/second-circuit-affirms-fair-use-in-google-books-case/.
Davis, Michelle. “Anthropic’s Great Book Heist: Do the Ends Justify the Means When It Comes to Training AI?” Commercial. JD Supra, August 27, 2025, https://www.jdsupra.com/legalnews/anthropic-s-great-book-heist-do-the-5275505/. Accessed 5 Mar. 2026.
Google NotebookLM - Copyright Questions | ACRL Artificial Intelligence (AI) Interest Group. https://connect.ala.org/acrl/discussion/google-notebooklm-copyright-questions. Accessed 5 Mar. 2026.
“Google Workspace Updates: NotebookLM Is Now Available to All Education Users.” Workspace Updates Blog, http://workspaceupdates.googleblog.com/2025/08/notebooklm-is-now-available-to-all.html. Accessed 5 Mar. 2026.
Lim, Alex. “Gemini Certified Educator: Is Your School’s Data Safe with NotebookLM? Understanding Google Powerful Privacy Guarantees for Education.” PUPUWEB, 10 Sept. 2025, https://pupuweb.com/gemini-certified-educator-is-your-schools-data-safe-with-notebooklm-understanding-google-powerful-privacy-guarantees-for-education/.
NotebookLM Help. Learn about NotebookLM - Computer - NotebookLM Help. https://support.google.com/notebooklm/answer/16164461?hl=en&ref_topic=16164070&sjid=15988338516415528594-NA. Accessed 5 Mar. 2026.
Workspace, Google. “NotebookLM: AI-Powered Research and Learning Assistant Tool.” Commercial. Google Workspace, Google, https://workspace.google.com/products/notebooklm/. Accessed 5 Mar. 2026.