From 26 October 2026, new paid-tier Gemini API projects store Generate Content requests and responses by default, matching the Interactions API. Existing projects keep their current setting. Teams sending personal or confidential data should decide now whether to keep stored logs, shorten retention, or opt out with store=false per request or per project.
What Google changed
The change was announced by Google on 26 October 2026 in the Gemini API changelog. New projects on the paid tier that call generateContent, streamGenerateContent or batchGenerateContent now have store=true by default, the same default the Interactions API already uses. Projects created before that date are not switched over.
Google points to two ways out: set store=false on individual requests, or turn off Generate Content API logging in the Settings panel of the Logs and Datasets dashboard.
What gets stored, and for how long
The logs and datasets guide sets out the details:
- Content: each log holds the prompt, the response and the context from previous turns. Interactions API logs also link to the previous interaction.
- Retention: logs expire after a default window set out in the logs and datasets guide, and you can choose a shorter one per project. Logs you save into a dataset do not expire.
- Paid tier only: storage is available only on billing-enabled projects, so free-tier projects are not affected.
- Not logged: Imagen and Veo, Gemini embedding models, the Gemini Robotics model, public preview agents, and requests containing video, GIFs or PDFs.
- Visibility: logs appear in AI Studio one project at a time, and only while the project has at least one active API key.
On data use, the data logging and sharing policy says logs are private to your Google Cloud project and that prompts and responses in logs are not used for product improvement by default. That changes if you share a dataset with Google: shared data may be read by human reviewers and used to improve models, under the unpaid services terms.
How to opt out, and the version you need
The request-level setting takes precedence over the project setting, so you can keep logging on for a project and switch it off for sensitive calls, or the reverse. The store parameter needs a recent SDK: google-genai 2.33.0 or later for Python, @google/genai 2.31.0 or later for JavaScript, google-genai 1.79.0 or later for Java, and google.golang.org/genai 1.76.0 or later for Go. Older clients cannot send it, so pin the version before relying on a per-request opt-out.
In the dashboard, Generate Content and Interactions logging are separate switches. Turning off Interactions logging also stops automatic storage of conversation history, so check whether any feature relies on server-side state before you flip it.
What it means for teams building AI products
CodeDTX's view: stored logs are useful and risky at the same time. They give you real traffic for evaluation sets and debugging without building your own capture pipeline. They also create a second copy of user data outside your own systems, with its own retention clock.
- Audit which projects are new. A project created this week behaves differently from one created last month. If your team creates a project per environment or per customer, the default now depends on the creation date.
- Treat it as a data processing decision. If prompts carry personal data, the stored copy falls under the same rules as your own logs. Our guide on keeping personal data out of AI agent logs covers what to redact first.
- Set retention deliberately. If you keep logging, choose the shortest window that still serves debugging, and record who can open AI Studio for that project.
- Be careful with datasets. Saved datasets never expire, and sharing one with Google moves it under different terms. Make dataset creation a reviewed step, not something any engineer does ad hoc.
- Check residency needs. The guide does not describe where logs are held. If you have data residency requirements, confirm them with Google before keeping storage on.
When not to change anything
If your projects predate 26 October 2026, nothing has changed for them. If you are on the free tier, storage is not available. And if you already run your own request logging with redaction and retention controls, turning Gemini storage off for production avoids a duplicate copy while you keep it on in development for evaluation. Teams also moving models, as covered in our post on Gemini 4 Argon API changes, can make both changes in one release.
Frequently asked questions
Does the change affect Gemini API projects created before 26 October 2026?
No. Google's changelog says existing projects keep their current setting, and the logs guide says Generate Content storage defaults to on only for projects created on or after that date. Older projects stay off unless someone enabled logging in AI Studio. It is still worth checking each project's Settings panel, because a teammate may have switched logging on earlier.
Is data stored by Gemini API logging used to train Google's models?
Not by default. Google's data logging policy says prompts and responses in logs are not used for product improvement or development. That changes when you share a dataset with Google, which you can opt out of when creating it. Shared data may be read by human reviewers and used to improve model quality, so avoid putting personal or confidential content into shared datasets.
How do I turn off Gemini API request storage?
You have two options. Set store to false on each request using a recent google-genai SDK, or turn off Generate Content API logging in the Settings panel of the AI Studio Logs and Datasets dashboard. The request setting overrides the project setting, so you can keep logging on in general and switch it off only for calls that carry sensitive data.
How long are Gemini API logs kept?
Logs are kept for a default retention window, which you can shorten per project in the Logs and Datasets dashboard, after which they expire and are marked for deletion. Logs you copy into a dataset do not expire, so a dataset needs its own clean-up rule. Removing every active API key from a project also hides its logs from view in AI Studio.
Should we keep Gemini request storage on?
It depends on what your prompts contain. Keeping it on gives you real traffic for evaluation sets and debugging with little effort. If prompts include personal or regulated data and you already log requests yourself with redaction, turning it off in production avoids a second copy. A practical split is to keep it on in development and test projects and off in production.
