BLUF: Google has renamed NotebookLM to Gemini Notebook and is expanding it from a source-grounded research workspace into a cross-product context layer with native code execution for eligible plans. For enterprise teams, the material change is not the name. It is that a notebook can now become a path by which files, prompts, outputs and analysis move across more surfaces. Treat that path as an integration boundary, not as a personal productivity feature.
What changed
Google announced that NotebookLM is now Gemini Notebook. It remains a standalone research product, but Google says notebooks can sync between the Gemini app and the standalone experience; notebooks are also planned for AI Mode in Search. Google has separately documented that notebooks in the Gemini app can hold chats, custom instructions and uploaded files, and that sources added in one product appear in the other.
The more consequential capability is a secure cloud computer for each notebook. Google says it enables native code writing and execution for data analysis grounded in notebook sources. At announcement time, this was available to Google AI Ultra users and Workspace business customers with AI Ultra Access and AI Expanded Access, with a planned web rollout to Pro users. Availability is therefore a commercial and tenant-configuration question, not an assumption for every employee.
Google announcement: NotebookLM is now Gemini Notebook
Google announcement: notebooks in the Gemini app
Why this changes the operating model
A conventional research assistant is relatively easy to contain: define approved sources, control who can access the workspace, and review its output before it leaves the team. Cross-app sync changes the asset boundary. A notebook can now carry a mixture of source files, conversation context and instructions into another interface with different sharing behaviour, retention expectations and user habits. Native code execution adds another boundary: generated analysis can be useful, but it must be reproducible and reviewable before it informs a decision.
This is not a reason to block the product. It is a reason to stop treating a notebook as an informal scratchpad. The appropriate unit of governance is the notebook and its connected surfaces: named business owner, approved purpose, source classification, authorised user group, allowed outputs and a deletion or offboarding path. The same discipline is useful for enterprise RAG systems: permissions must be enforced before retrieval, and the system needs evidence of its access decision. See Ade’s practical guide to RAG access control: https://adechristanto.de/en/blog/rag-access-control-enforce-permissions-before-retrieval.
Data residency and procurement: verify the exact service boundary
German and European buyers should not infer data residency from a Workspace contract alone. Google’s Workspace Service Specific Terms say that data-region commitments apply to defined “Located Data” for in-scope editions, and that customer data outside that policy may be stored or processed wherever Google or its subprocessors have facilities. The list explicitly includes Gemini in Workspace prompts and generated output, but procurement must establish whether the particular Gemini Notebook capability, plan and processing path are covered by the customer’s contract and configuration.
This is a procurement control, not legal advice. Ask the vendor for the current product-specific terms, data-processing documentation, subprocessor position, feature availability by edition, and the status of any preview capability. Do not load special-category data, confidential customer material, export-controlled information or security-sensitive incident evidence into a new feature until the data owner has approved that exact path. The difference between a generally available service and a preview matters: Google’s terms allow use of customer test data to provide, test, analyse, develop and improve pre-GA offerings, subject to the stated terms.
Google Workspace Service Specific Terms
The technical limits are still real
“Grounded in sources” does not make an output correct. Google’s own help documentation for generated slide decks warns that they may contain visual or factual inaccuracies. Code generated inside a notebook can also produce plausible but wrong transformations, choose an unsuitable statistical method, or hide assumptions in generated cells. Source sync does not repair stale permissions; it can make stale permissions travel farther.
Apply the same production controls used for agentic automation. Keep a fixed source set for high-stakes work. Require a human reviewer for external reports, financial analysis, operational changes and any action based on generated code. Preserve the source version, prompt, output and reviewer decision. Define an escalation route for inaccessible, revoked or disputed sources. These controls follow the approval-gated pattern described in https://adechristanto.de/en/blog/human-in-the-loop-ai-agents-approval-logs-evals-2026 and make later review possible rather than theoretical.
A practical decision: pilot one bounded notebook
Start with a non-sensitive, repeatable research workflow: for example, summarising approved product documentation for an internal enablement team. Create one controlled notebook; set the source owner and reviewer; disable public sharing where it is not needed; test sync into each enabled Gemini surface; and run a deliberately revoked-file test. For code execution, require a saved input set, a peer review of the generated logic and an independently reproducible result before decisions depend on it.
Measure more than time saved: track source-authorisation failures, unsupported claims found in review, output corrections, permission-revocation latency and the proportion of outputs with complete evidence. These are the operating measures that turn a convenient research capability into an accountable business process. Ade’s guide to production AI evaluation provides a useful metric baseline: https://adechristanto.de/en/blog/ai-agent-evaluation-production-cost-accuracy-escalation-2026.
The next enterprise question
Gemini Notebook can reduce the friction between research, analysis and drafting. That also reduces the friction between a trusted document repository and an unreviewed AI interaction. Teams that define the notebook as a governed integration object now will be able to use the capability faster when access expands, without discovering their data, permission and evidence gaps in a live project.
Primary sources
Google: NotebookLM is now Gemini Notebook
Google: Try notebooks in Gemini to easily keep track of projects
