PrivaSearch

AI and privacy

NotebookLM and the GDPR: can you put business documents in it?

NotebookLM is now Gemini Notebook. What happens to uploaded sources for each account type, how sharing works and what this means for client files.

In brief

  • Since 16 July 2026, NotebookLM has been called Gemini Notebook: the same product, with the same notebooks and sources.
  • Every source you add goes to Google in full, even if you only have one question about one paragraph.
  • A personal account is subject to Google's general terms and has no data processing agreement.
  • For Workspace accounts, Google says uploads, questions and answers are not reviewed by humans or used for training (as of October 2026).
  • According to Google, your organisation's file sharing and data region settings do not apply to data in the notebook.

Can you put business documents in NotebookLM?

On a personal Google account, do not put documents containing personal data or confidential information in NotebookLM: there is no data processing agreement and your organisation has no visibility into what is shared. A Workspace account covered by the business agreement can be used if you have assessed the settings, sharing and data location. The entire document is always included.

This article describes Google's own documentation, consulted on 13 October 2026. It is an explanation, not legal advice: assessing your own client files remains the responsibility of you and your data protection officer.

First, the name: NotebookLM is now Gemini Notebook

On 16 July 2026, Google announced that NotebookLM would become Gemini Notebook. According to the announcement, it remains the same standalone product, with more connections to the rest of Google. The help pages now use the new name throughout, sometimes followed by “formerly NotebookLM”.

This article uses NotebookLM when discussing the product generally, because most people still use that name. When discussing the terms, it uses Gemini Notebook, because that is the name you will find in the documents.

What is NotebookLM and how does it work?

Google describes it on the product help page as an AI-powered research assistant. You create a notebook, add sources and ask questions. The answer comes from those sources and references them.

It is RAG on your own files

Technically, this is an application of retrieval-augmented generation: the system first finds the relevant parts of your sources and lets the model answer using them. This allows you to check an answer against the source. How that technology works and where it falls short is explained in RAG explained.

The technology does not change one thing under the GDPR. To search your document, the provider must first have it. For what this means for chatbots generally, see Uploading documents to an AI chatbot.

Audio overviews

According to the help page on audio overviews, an audio overview is a conversation between AI voices summarising the main topics in your sources, in more than eighty languages. Google warns that an overview may contain inaccuracies. You can download it as a file or share it using a link. In interactive mode, you can ask questions using your voice; according to Google, that voice and its transcript are not stored or shared.

An audio file about a client file is a new document containing the same personal data, in a format that is easy to circulate and that nobody can search. The section on client files returns to this.

What happens to your sources on a personal account?

A personal Google account is subject to Google's general terms. The help page makes three points about this.

Training and feedback

According to Google, your data is not used to train Gemini Notebook unless you provide feedback. If you submit feedback, Google can review the entire context of that interaction: your questions, uploads and the model's answers. Giving a thumbs up or down to an answer about a client file is therefore a moment when that file could reach a reviewer.

Google also writes that it may process your data to prevent fraud, abuse and technical problems. The help page explicitly promises Workspace users that nobody reads along. For personal accounts, that statement is not made in the same form.

Notebooks in the Gemini app

Notebooks also appear in the Gemini app, with changes synced between the two. The help page on notebooks in the Gemini apps says conversations you have with a notebook there are retained according to your activity setting and used as described on the Gemini apps' privacy page. On a personal account, this can mean training and human review. What that setting involves is explained in Google Gemini and the GDPR.

The same source can therefore fall under two sets of rules, depending on the window in which you ask the question. In the Gemini app, the answer can also draw on web search results, rather than just your sources.

No agreement with your organisation

A personal account has no data processing agreement of the kind required by Article 28 of the GDPR. Your organisation cannot configure anything, inspect anything or delete anything. An employee putting work documents in a private notebook operates outside your visibility. How to address this is explained in Shadow AI.

What changes with a Workspace account?

For most Google Workspace editions, Gemini Notebook is a core service. According to the Gemini Notebook administration page, this includes Business, Enterprise and Frontline editions, as well as education. For some other account types, it is an additional service. An administrator turns it on or off for each group of users.

The agreement and the commitments

The Generative AI in Workspace Privacy Hub says Gemini Notebook is covered by your organisation's Workspace agreement, including the data processing addendum. According to that page, uploads, questions and answers are not reviewed by humans or used to train generative AI models.

That page applies to eligible Workspace editions. It explicitly excludes Gemini Notebook for education, which has its own pages. For what else to check in such an agreement, see A data processing agreement with an AI provider.

What is retained

That last point is easily missed. Do not assume the notebook follows when you delete a file or revoke permissions in Drive: check in the notebook itself.

Three settings that do not carry over

The Privacy Hub identifies three limitations administrators need to know, as of October 2026:

Carefully configuring what may leave your domain and where data is held in Drive therefore does not settle this for notebooks. Read the addendum itself to establish whether and how data reaches a country outside the EEA. Record the outcome in your record of processing activities.

The version through Google's cloud service

Alongside Workspace, there is Gemini Notebook Enterprise, which organisations obtain through Google's cloud service. According to the documentation, source data is held in the client's own project, in the United States or the EU; notebooks cannot be shared publicly and may only be shared with users within the same project. This version is subject to that cloud service's terms. The overview page says nothing about training; look this up in the agreement before making assumptions.

What happens when you share a notebook?

Sharing is where even a carefully chosen source can travel too far.

Sharing with people

According to the help page on creating and sharing notebooks, you share by adding an email address with one of two roles:

Two details deserve attention. If you allow copying, a recipient can duplicate the notebook, including its sources. And the view that shows a viewer only the chat hides the sources but does not revoke access: Google itself writes that viewers may still be able to access hidden materials.

According to the help page on public notebooks, owners and editors can open a notebook to anyone with the link and a Google account. Such visitors can ask questions and view the generated overviews, and may also be able to access the sources. Access can be revoked in the sharing panel; the old link then stops working.

The help page on audio overviews says public sharing is available only for consumer accounts and is disabled for Workspace Enterprise and education accounts. Check what a user can do in your own environment before making assumptions.

Accounts side by side

As of October 2026, based on Google's documentation. “Suitable for” is an initial classification, not a judgement on your own processing.

What does this mean for client files containing personal data?

A client file rarely belongs solely to your organisation. It contains clients, employees or patients who do not know their data is ending up in a notebook.

Four points that apply to every account

An example

A fictional bookkeeping firm, Kantoor Voorbeeld, wants faster answers about clients' employment arrangements. An employee puts twenty employment contracts in a notebook on a personal account and shares it with two colleagues as editors. One creates an audio overview and forwards the file.

There are now twenty contracts with names, salaries and addresses held by a provider without an agreement, three people who can share the notebook further, and an audio file nobody can retrieve. None of this is in the record of processing activities.

What to arrange before your team starts

  1. Choose the account. Work accounts only, and check whether Gemini Notebook is a core service in your edition.
  2. Decide who may use it. The administrator can enable the service for each group.
  3. Agree which sources may be included. By document type, with examples. See Creating an AI policy for your organisation.
  4. Set sharing rules. Who may be an editor, whether copying is allowed, and that public sharing is prohibited.
  5. Treat derived files like the client file. Including the audio overview.
  6. Clean up. Document when notebooks and sources are deleted and who does this.
  7. Record it. An entry in the record of processing activities, and a DPIA if the processing is likely to involve a high risk. See DPIAs for AI tools.

What PrivaSearch does here

NotebookLM and PrivaSearch both answer questions from your own documents with a source reference. The difference lies in what happens beforehand. Inside PrivaSearch, personal data is detected and replaced with consistent pseudonyms, and a colleague checks the result before a document becomes searchable. The steps are explained on How PrivaSearch works.

Search takes place per project, and only for people with access. You can search entirely within PrivaSearch. Semantic search then uses a European AI provider, and only text that has been pseudonymised and checked. Processing and storage take place in data centres in the EU. The principles are on Hosting in the EU, and the parties involved are in the sub-processor overview.

Those wanting an AI assistant to help search can connect Claude, ChatGPT Business or Mistral Vibe (formerly Le Chat). The assistant then receives only released, pseudonymised passages and no original files. What goes to the assistant's provider in that process is explained on the page about AI chatbots and the GDPR.

Pseudonymisation is not anonymisation: for your own organisation, this remains personal data. The lawful basis, the DPIA and the assessment of your client files remain the responsibility of you and your data protection officer.

Want to see how a question about a client file is answered without sending the file along? Book a demo.

Frequently asked questions

Is NotebookLM safe for business documents?

That depends on the account and what the documents contain. A personal Google account has no data processing agreement and your organisation has no visibility into what is shared. With a Workspace account, the product is covered by the business agreement, without human review or training. Even then, the whole document goes to Google and you must assess sharing and data location yourself.

Does Google train on the sources I upload to NotebookLM?

According to the help page, your data is not used to train the product unless you provide feedback. With feedback, Google can review the entire context: your questions, uploads and the answers. For Workspace accounts, Google says uploads, questions and answers are not reviewed by humans or used for training. This is the position as of October 2026; read the help page yourself.

Is NotebookLM now called Gemini Notebook?

Yes. Google renamed NotebookLM to Gemini Notebook on 16 July 2026. According to the announcement, it remains the same standalone product, with more connections to the Gemini app and other Google services. The help pages and administration documentation now use the new name. When looking up the terms, use both names.

Can someone else see my notebook?

Only if you share it. You share with people as viewers or editors, and an editor can share the notebook further. On a personal account, you can also open a notebook to anyone with the link and a Google account. According to Google, a viewer shown only the chat may still be able to access the sources. Only share with people who may see the sources themselves.

Where are the files I put in NotebookLM stored?

At Google, with the notebook's data. If you add a file from Drive, Google says the product creates a new copy that is not held in Drive. According to Google, the data region an organisation has chosen for Workspace does not apply to data in the notebook. Only with the version through Google's cloud service does an administrator choose between storage in the US and the EU.

May I put an employee file in NotebookLM?

Not on a personal account: there is no agreement with your organisation and nothing has been documented. With a Workspace account, it depends on your lawful basis, the type of data, sharing rules and your own assessment, preferably in a DPIA. Remember that the whole file is included and that a summary or audio overview contains the same data. Ask a lawyer to review it if in doubt.

Read more

  1. Uploading documents to ChatGPT or Claude: what happens to them?

    Where an uploaded file ends up, how long it stays, who can access it and what deletion means. With four approaches for files containing personal data.

  2. AI in education: may a school put pupil data into an AI chatbot?

    How schools use AI, why pupil data receives extra protection, which applications are prohibited or high risk, and what a school needs to arrange.

  3. Is ChatGPT safe? What happens to your data and which settings you control

    What ChatGPT stores, who can access it, how to turn off training, and what temporary chats, memory and shared links do. Includes a practical checklist.

See in twenty minutes what happens to your documents.

We show it with fictitious client files from your sector: upload, detection, review, search and the connection with your AI assistant. Afterwards you'll know exactly what your team can use safely.

Book a demo →