Careful with sensitive material

Never put this into it

NotebookLM is built to work with your own uploaded sources, which makes it tempting to upload a whole folder of interview transcripts or case files. Avoid doing that on a personal Google account. Do not upload files containing witness names, exact locations, or ID numbers unless your organisation is on the free Google Workspace for Nonprofits programme or a paid Workspace/Gemini Enterprise plan, and never leave a thumbs-up/down rating on a response that quotes sensitive material. Google says feedback can be used to improve its models even when the underlying source upload is not.

A safer tool to use instead

NotebookLM, which Google has renamed Gemini Notebook, answers questions using only the documents you give it. Upload a set of reports, transcripts or draft legislation, ask across all of them at once, and each answer points back to the passage it came from.

That grounding makes an invented answer much less likely than from a general chatbot, which is exactly the trap. It feels safe enough to upload the whole folder. The restriction is on where its answers come from, not on where your files go.

Registered nonprofits can get it free, with contractual data protection, through Google’s nonprofit programme. That is a better starting point than a personal Google account.

What happens to your data

Where this data goes

  • Google servers (multiple countries)

Trains on what you type?

No. Your input is not used to train models.

How long it is kept

Google states uploaded sources are not used to train its generative AI models, and are only visible to you and anyone you explicitly share a notebook with. The one exception: Google says feedback you actively submit (a thumbs-up/down on a response) can be used to improve its models, so a rating left on a response that quoted sensitive material could expose that quoted text. Enterprise and Google Workspace for Nonprofits accounts carry a contractual commitment that documents and outputs are never used to train models, with no feedback exception.

Account required

  • An email address

Getting it, and using it in Iraq

Access in Iraq

Works in Iraq
  • Paid plans typically need a card issued outside Iraq.

Google states NotebookLM is available in the 180+ regions where the Gemini app is available, and Iraq is on Google's own published Gemini-app country list, so it is reasonable to expect access, though Google does not publish a NotebookLM-specific country-by-country list to check this against directly. As with other Google paid plans, organisations in the region report Iraqi-issued cards are often declined for the paid Google AI plans that raise its limits; the free tier and nonprofit programme above need no card.

Language quality

No one has published a measurement for this.

  • Arabic (Modern Standard)

    Google states NotebookLM supports chat and sources in dozens of languages and lists Arabic among its interface languages, but publishes no per-language accuracy figure, and no independent benchmark was run. Untested.

  • Iraqi Arabic

    No dialect-specific benchmark exists or was checked for this entry. Untested.

  • Kurdish (Sorani)

    No benchmark or vendor documentation on Kurdish Sorani support was found or checked for this entry. Untested.

Sources

Last checked: 2026-08-31