It depends on which Gemini you are signed into. Google's Gemini Apps Privacy Notice says that on a personal Google account, activity is used to provide, develop and improve its services, including training generative AI models, and that human reviewers review some chats. Google's Workspace privacy hub says that on a qualifying Workspace edition, content is not human reviewed or used for generative AI model training outside your domain without permission. Google Cloud and the Gemini API have their own terms again. The word "Workspace" alone does not settle it, because Google applies the personal-account notice to users without a qualifying edition. This is general information, not legal advice.
Reviewed and corrected on 12 October 2026. Every Google statement below comes from the Google page named next to it, which we read on that date. Google changes these terms often, so check the pages before relying on them.
TL;DR
- A personal Gemini account is not the setup to use for client data. Google's notice says human reviewers review some chats and asks users not to enter confidential information they would not want a reviewer to see or Google to use to improve its services. It also says not to rely on Gemini Apps for "legal, financial, or other professional help."
- Turning Keep Activity off narrows the exposure and does not end it. Google says future chats are still saved for 72 hours, and that even with the setting off it still uses chats to respond to you and protect Google and its users, including with help from human reviewers.
- Chats already reviewed by people are kept up to three years. Google says they are disconnected from your account and are not deleted when you delete your activity.
- Workspace is only protected on a qualifying edition. Google's admin help page says users without a qualifying edition are subject to the Google Terms of Service and the Gemini Apps Privacy Notice, and their chats may be reviewed by human reviewers.
- Google Cloud, the Gemini API and Workspace each have separate terms. The free-of-charge API tier and the paid tier are treated differently.
- ABA Formal Opinion 512 still applies. A vendor's data terms help you decide, but you still have to read them and weigh consent.
Your firm pays for Google Workspace. A partner opens Gemini and pastes a client's draft agreement. Which fact tells you whose data terms apply?
Which Gemini are you on?
People say "Gemini" as if it were one product. For confidentiality it splits at least four ways, and each carries different terms.
- Gemini Apps on a personal Google account. The app at gemini.google.com and the mobile apps, signed in with a personal account. Governed by the Gemini Apps Privacy Notice.
- The Gemini app and Gemini in Workspace on a qualifying Workspace edition. Governed by the Generative AI in Google Workspace Privacy Hub and the Cloud Data Processing Addendum.
- Google Cloud. Two documents matter. One covers the model platform that Google's documentation now calls Gemini Enterprise Agent Platform (Google's name-changes page says some Vertex AI products and features have been renamed, and the old Vertex AI data governance address now lands on the new page). The other covers Gemini for Google Cloud, the assistance inside the Cloud console and developer tools.
- The Gemini API. For teams building their own tools, under the Gemini API Additional Terms of Service, which split into unpaid and paid services.
A lawyer who opens gemini.google.com with a Gmail login is on the first one.
Personal Gemini versus Workspace, side by side
All entries below are Google's statements, from the pages named in the Sources section, as of 12 October 2026.
| Question | Personal Gemini account | Qualifying Workspace edition |
|---|---|---|
| Used to train Google's models | With Keep Activity on, Google uses activity to provide, develop and improve its services, including training generative AI models | Google says content is not used for generative AI model training outside your domain without permission, citing the Training Restriction in the Workspace Service Specific Terms |
| Human review | Google says human reviewers, including trained reviewers from its service providers, review some of the data | Google says content is not human reviewed outside your domain without permission |
| Retention | Auto-delete defaults to 18 months (3 months, 36 months or indefinite are the options); chats reviewed by people are kept up to 3 years; with Keep Activity off, chats are saved for 72 hours | Set by administrators: Gemini in Workspace prompts and responses 90 days to indefinite; Gemini app up to 36 months, with 18 months the default |
| Terms and controls | Google Terms of Service and the Gemini Apps Privacy Notice | The Cloud Data Processing Addendum, the Workspace terms, and administrator controls |
| What to confirm before client data goes in | Do not use it for client matters | That your edition qualifies, what your administrator has set, and who can access the account |
Here is the same logic as a decision path.
What Google says about the personal Gemini app
Google's notice says: "Human reviewers (including trained reviewers from our service providers) review some of the data we collect for these purposes. Please don't enter confidential information that you wouldn't want a reviewer to see or Google to use to improve our services, including machine-learning technologies." In its list of other things to know, it adds: "Don't rely on Gemini Apps for diagnosis, treatment, medical advice, or legal, financial, or other professional help."
Three more points from the same notice (notice dated June 29, 2026; hub page last updated September 24, 2026):
- Keep Activity on. Google says it uses your activity to provide, develop and improve its services, including training generative AI models, and to protect Google, its users and the public with the help of human reviewers.
- Reviewed chats are kept. Chats reviewed by human reviewers, and related data such as language, device type, location and feedback, are not deleted when you delete your activity. Google says they are retained for up to three years, disconnected from your Google Account.
- Keep Activity off, or a Temporary Chat. Google says temporary chats are not used to train its AI models, and that with Keep Activity off, future chats are not used to train them unless you choose to send feedback. Both are still retained with your account for 72 hours. Google says it needs that time to respond, to protect Google and its users, and to process any feedback. If you send feedback, Google says it also collects the last 24 hours of your chats and any content in them, and reviewed feedback is retained for up to three years.
The result is that the off switch and Temporary Chat reduce what feeds model improvement, but Google's own text says it still uses chats for protection purposes, including with human reviewers. A setting you control is not a signed agreement.
Google's notice also does not carve out paid personal Google AI plans. It lists subscription information among the data it collects. We found no separate data terms for paid personal plans on the pages we read, so do not assume a paid personal subscription changes the picture. Not verified in our October 2026 review. Check the notice and your plan's terms.
What Google says about Workspace
The Generative AI in Google Workspace Privacy Hub (last updated October 6, 2026) says that Gemini interactions stay within your organization, that Gemini does not share your content outside it without permission, and that "Your content is not human reviewed or otherwise used for Generative AI model training outside your domain without permission." It also says user prompts are customer data under the Cloud Data Processing Addendum and that Workspace does not use customer data to train models without the customer's prior permission or instruction.
That hub applies only on a qualifying edition. Google's help page "Turn the Gemini app on or off" (last updated October 9, 2026) lists the editions that include the Gemini app as a core service with enterprise-grade data protections: AI Expanded Access, Business Starter, Business Standard, Business Plus, Enterprise Starter, Enterprise Standard, Enterprise Plus, Essentials, Enterprise Essentials, Enterprise Essentials Plus, Frontline Starter, Frontline Standard, Frontline Plus and Nonprofits. The page lists Workspace for Education editions in a separate section. For the Business editions it notes that the Gemini mobile app requires signing up with a business email address or verifying your domain. It says users without a qualifying edition are subject to the Google Terms of Service and the Gemini Apps Privacy Notice, and that their chats may be reviewed by human reviewers.
A few more Workspace details from the hub:
- Retention is an administrator setting. The hub's table gives 90 days to indefinite for Gemini in Workspace prompts and responses, and up to 36 months for the Gemini app. For the app, administrators can choose 3, 18 or 36 months, with 18 months the default, and when conversation history is off, new chats are saved in user accounts for up to 72 hours.
- Feedback is its own channel. If a user chooses to submit feedback, it can include the prompt, contextual sources and output. Google says feedback may be retained up to 18 months and is not used to train the models behind Workspace's generative AI services.
- Gemini only retrieves what the user can access, and restrictions such as disabling download, copy or print, and client-side encryption, can keep files out of its reach. Google also says Data Loss Prevention applies to generated output inserted into supported Workspace services for supported editions.
So the Workspace protection is real on paper, but it depends on three things: the edition, your administrator's settings, and the contract. Ask for the edition and the settings in writing.
What Google says about Google Cloud
Google Cloud's page "Gemini Enterprise Agent Platform and zero data retention" (last updated October 9, 2026) says that, under the Training Restriction in the Service Specific Terms, Google "won't use your data to train or fine-tune any AI/ML models without your prior permission or instruction," and that this applies to all managed models on the platform, including GA and pre-GA models. The same page lists the cases where data is retained for limited periods, which matter if you need zero data retention:
- Abuse monitoring. Google may log prompts to detect abuse. This applies to customers whose Google Cloud use is governed by the Google Cloud Platform Terms of Service, and a customer in scope who wants zero data retention can request an exception. Some advanced AI models or features may log prompts and responses for abuse prevention, and zero data retention may not be possible with them.
- Grounding with Google Search. Google stores logs of queries derived from end-user prompts for up to three days and says there is no way to disable that storage if you use the feature.
- In-memory caching. Google's published Gemini models cache customer data in memory with a 24-hour time to live, isolated at the project level. The page says this does not violate zero data retention and can be disabled at the project level.
A separate Google page, "How Gemini products in Google Cloud use your data" (last updated October 9, 2026), covers Gemini for Google Cloud, the assistance in the console and developer tools. It says: "Gemini doesn't use your prompts or its responses as data to train its models."
What Google says about the Gemini API
The Gemini API Additional Terms of Service (last updated April 28, 2026) split into two. For Unpaid Services, which include "direct interactions with Google AI Studio or unpaid quota in Gemini API," Google "uses the content you submit to the Services and any generated responses to provide, improve, and develop Google products and services and machine learning technologies," and "human reviewers may read, annotate, and process your API input and output." The terms add: "Do not submit sensitive, confidential, or personal information to the Unpaid Services." For users in the European Economic Area, Switzerland and the United Kingdom, the paid-services data terms apply to all services.
For Paid Services, Google says it "doesn't use your prompts ... or responses to improve our products," processes them under the applicable Data Processing Addendum, and "logs prompts and responses for a limited period of time, solely for detecting and preventing violations of the Prohibited Use Policy" and for any required legal or regulatory disclosures. The terms say Google AI Studio access counts as a Paid Service, even when free, if the account has access to a Cloud project with active billing or is a Workspace enterprise account. If your developers build on the API, the tier decides the training posture, so find out which one your key is on.
Safe and unsafe lawyer uses
The duty of confidentiality does not pause for a new tool. ABA Formal Opinion 512, an advisory ethics opinion, says lawyers should read and understand the terms of use and privacy policy of any generative AI tool they use, or consult someone who has. It says a client's informed consent is required before inputting information relating to the representation into a self-learning tool, that general boilerplate in an engagement letter is not enough, and that consent is not needed where the lawyer does not input such information, for example idea generation that uses no client facts.
Whether pasting privileged material into a given tool puts privilege at risk is a separate question, and we have not verified any decision about Gemini. For a court's reading of a different chatbot's privacy policy, see our Claude privacy guide.
Reasonable on a personal account (no client information):
- Drafting a plain-English explainer of a public legal concept.
- Rewriting or shortening text that contains no client or matter detail.
- Brainstorming an outline using hypothetical, anonymized facts.
- Summarizing a published statute or public court opinion that you then read yourself.
Not appropriate on a personal account:
- Pasting a client contract, term sheet or settlement figure.
- Entering names, parties or any detail that identifies a matter.
- Uploading privileged communications or work product.
- Typing anything you would not want a Google reviewer to read.
The dividing line is simple. If a reviewer reading it would breach a confidence, it does not belong in a personal Gemini account. On a qualifying Workspace edition or a Google Cloud agreement, the calculus changes, but you still verify the edition, the settings and the terms, and you match the tool to the sensitivity of the matter. Our guides on ChatGPT confidentiality and Claude for legal work walk through the same plan-versus-terms split for those tools, and where your legal AI data actually goes covers where inputs travel across tools.
The verdict
A personal Gemini account is not the setup to use for client-confidential work, and Google's own notice describes why: human reviewers read some chats, reviewed chats are kept for up to three years, and the off switch does not end Google's own use of chats for protection. A qualifying Workspace edition, a Google Cloud agreement or a paid API tier comes with different written commitments, but they depend on the edition, the product and the administrator's settings, so confirm each of those and read the current terms. Where client information is involved, handle consent under ABA Formal Opinion 512.
For the US statutes, regulations and court rules these laws sit in, see the US primary law API.
Sources, all read on 12 October 2026: Google, Gemini Apps Privacy Hub (hub last updated September 24, 2026; Gemini Apps Privacy Notice dated June 29, 2026). Google, Generative AI in Google Workspace Privacy Hub (last updated October 6, 2026). Google, Turn the Gemini app on or off (last updated October 9, 2026). Google Cloud, Gemini Enterprise Agent Platform and zero data retention (last updated October 9, 2026) and Name changes from Vertex AI (last updated October 9, 2026). Google Cloud, How Gemini products in Google Cloud use your data (last updated October 9, 2026). Google, Gemini API Additional Terms of Service (last updated April 28, 2026). American Bar Association, Formal Opinion 512 (July 29, 2024).
FAQ
Is Google Gemini private for legal work? Not on a personal account. Google says human reviewers review some chats there and that activity is used to improve its services, including training generative AI models. On a qualifying Workspace edition, Google says content is not human reviewed or used for model training outside your domain without permission. You still need to confirm the edition and settings and read the terms.
Does Gemini train on what I type? On a personal account with Keep Activity on, Google says it uses your activity to provide, develop and improve its services, including training generative AI models. With Keep Activity off, Google says future chats are not used to train its AI models unless you choose to send feedback. Check the current notice, since these terms change.
Do humans read my Gemini chats? On a personal account, yes for a subset. Google says human reviewers, including trained reviewers from its service providers, review some of the data, and that chats reviewed this way are retained for up to three years, disconnected from your account. Google even says that with Keep Activity off it still uses chats to protect Google and its users, including with help from human reviewers.
Is my Google Workspace account covered by the stronger terms? Only if your edition is one Google lists as qualifying. Google's admin help page says users without a qualifying edition are subject to the Google Terms of Service and the Gemini Apps Privacy Notice. Ask your administrator which edition you have.
If I turn off Keep Activity, is my data deleted? No. Google says future chats are still saved for 72 hours, and chats already reviewed by people are retained for up to three years and are not deleted when you delete your activity.
Is Temporary Chat private? Google says temporary chats are not used to train its AI models. They are still retained with your account for 72 hours, and Google says it still uses chats to protect Google and its users. Temporary Chat is a mode, not a signed commitment.
How long does Gemini keep my conversations? On a personal account, Keep Activity auto-deletes after 18 months by default, with options of 3 months, 36 months or indefinitely. On Workspace, administrators set retention: Google's hub gives 90 days to indefinite for Gemini in Workspace and up to 36 months for the Gemini app.
Does the free Gemini API train on my data? Under Google's terms, Unpaid Services let Google use submitted content and responses to improve its products and allow human reviewers to read input and output, and the terms say not to submit confidential information to them. Paid Services are not used to improve its products, though prompts and responses are logged for a limited period for abuse detection. EEA, Swiss and UK users get the paid-services data terms on all services.
Does using Gemini waive attorney-client privilege? We have not verified a decision on Gemini. Privilege protects communications a client keeps confidential, so any tool whose terms let humans read your chats deserves caution, and the safe rule is to keep privileged material out of any tool whose terms you have not confirmed. This is general information, not legal advice.
Can I use Gemini and stay compliant with ABA 512? With care. Read the tool's terms of use and privacy policy, or have someone who has. Get the client's informed consent before inputting information relating to the representation into a self-learning tool; boilerplate in an engagement letter is not enough, and consent is not needed where no such information is input.
New legal AI guides, weekly.
Further Reading
Co-Founder & CEO · Attorney
Arshita leads product and strategy at Vaquill AI, the US primary law data API. As an attorney, she decides which law we carry, how each section is sourced and cited, and what legal AI and compliance teams need from the data.