Google turns on Gemini API request storage by default for new paid projects using the Generate Content API

Google has changed how the Gemini API treats request data for new paid-tier projects. According to an entry dated October 26, 2026 in the Gemini API release notes, new paid-tier projects that call the Generate Content API now store requests by default, which aligns that API with the behaviour already in place for the newer Interactions API. The change covers the three Generate Content methods, `generateContent`, `streamGenerateContent` and `batchGenerateContent`, and it means that prompts and responses sent from a newly created paid project will appear in the Logs and Datasets dashboard of Google AI Studio unless the developer opts out.

The release note is short, and its most important qualifications are in its second half. Existing projects keep their current setting, so teams that have been running production workloads on the Gemini API for months will not see their logging behaviour change overnight. Developers who do not want requests stored can switch storage off for an individual call by setting the `store` parameter to `false`, or turn Generate Content API logging off for the whole project in the Settings panel of the Logs and Datasets dashboard.

Until this update, the two main Gemini API surfaces had opposite defaults. Google's logs documentation states that the Interactions API stores all interaction objects with `store=true` by default, while the Generate Content API did not store requests unless storage was enabled per request or at project level in AI Studio. The October 26 change removes that asymmetry for new paid projects, and it moves the most widely used endpoint of the Gemini API from an opt-in model of logging to an opt-out one.

··········

⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣

WHAT THE OCTOBER 26 CHANGE COVERS.

The change applies to the logging feature that Google calls Gemini API logs. In Google's data logging and sharing policy, these logs are described as developer-owned data generated by supported Gemini API calls in projects with billing enabled, covering the path from request to model response, including requests, responses and metadata. Google states that the logs are private to the developer's Google Cloud project and are separate from the logs it keeps solely for abuse monitoring. In the AI Studio dashboard, a stored log shows the full prompt and response and, for multi-turn use, the context of previous turns.

Storage is limited to the paid tier. The logs documentation states that storage for Gemini API logs is only available for projects on the Gemini API paid tier, which is why the release note speaks specifically of new paid-tier projects. Free-tier projects are governed by different data-use terms, discussed below, and are not the subject of this default.

The scope of supported calls is broad. Google lists all `GenerateContent`, `BatchGenerateContent` and `StreamGenerateContent` calls, Interactions API calls with the exception of Managed Agents, and calls made through the OpenAI compatibility endpoints. That last point matters for teams that reach Gemini models through OpenAI-style client libraries: those requests are also eligible for logging and therefore fall under the same defaults.

........

Surface	Default before Oct 26	Default for new paid projects now	Existing projects
Generate Content API	Not stored	Stored	Keep current setting
Interactions API	Stored (store=true)	Stored (store=true)	Unchanged
OpenAI compatibility endpoints	Follow the logged API	Follow the logged API	Unchanged

........

··········

⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁤

WHAT CHANGES FOR GEMINI API DEVELOPERS.

For a developer who creates a new billing-enabled project after the change, the practical effect is that every supported Generate Content request will, by default, leave a copy of its prompt and response in the project's logs. That copy can be inspected in AI Studio, filtered, and assembled into datasets. For debugging, evaluation and prompt iteration, this is useful, because it gives teams a record of real traffic without extra instrumentation. For applications that process customer messages, personal data or confidential documents, it also means that a retention decision is now being made by default rather than by explicit choice.

Teams that already operate under internal data-handling rules should therefore review how new projects are provisioned. A company that creates a fresh project for each environment, customer or product line will find that each new project starts with Generate Content storage on, while older projects keep whatever setting they had. Google's release note does not specify the exact cutoff used to classify a project as new, so the safest assumption is that any project created from October 26 onward should be checked in the Logs and Datasets settings before it receives production traffic.

The two opt-out paths work at different levels. The project-level switch in the AI Studio Settings panel turns off logging for the Generate Content API for the whole project, and Google's documentation notes that logging can be toggled independently for `generateContent` and for the Interactions API. The per-request `store` parameter overrides behaviour for a single call. The documentation for the Interactions API adds a warning that is relevant to anyone who turns everything off: disabling Interactions API logging at project level stops the automatic storage and retrieval of conversation history, unless storage is re-enabled per request.

Logging also depends on API keys. The logs documentation says that a project needs at least one active API key for logs to be visible, and that deleting every key removes visibility into the project's logs until a key exists again.

··········

[[ADPLUS_300x250_1]]

HOW TO TURN STORAGE OFF PER REQUEST.

The logs documentation shows the per-request control in each of the official SDKs. In the Python example for the Generate Content API, the setting is passed in the request configuration as `config={'store': False}`, with the comment "Set to True to enable logging of this request"; the JavaScript equivalent uses `store: false`. The snippet below is adapted from that documentation example and shows the pattern with the Google Gen AI Python SDK; the model name is a placeholder.

from google import genai

client = genai.Client()

response = client.models.generate_content(
    model="MODEL_NAME",
    contents="Summarize this support ticket.",
    config={'store': False}  # Set to True to enable logging of this request
)

For the Interactions API, the same documentation passes `store=True` in Python and `store: true` in JavaScript, with a comment noting that the value can be set to false to disable logging of the request. Turning storage off there has functional consequences that do not apply to Generate Content. Google's Interactions API overview states that `store=false` is incompatible with background execution and prevents the use of `previous_interaction_id` in later turns, so an application that disables storage has to send the full conversation history itself.

··········

RETENTION, DATASETS AND GOOGLE'S USE OF THE DATA.

Stored logs do not persist indefinitely. Google's documentation sets a default retention window of 55 days, after which logs expire and are marked for deletion, and lets developers shorten the window to 7, 14 or 28 days. The data logging policy does not state how long deletion takes after a log is marked, which is a detail compliance teams may want to raise with Google. Logs that a developer adds to a dataset are treated differently: they do not expire and are kept beyond the retention window for downstream use.

The question most developers will ask is whether stored logs are used to train Google's models. Google's data logging and sharing policy states that, by default, prompts and responses in logs are not used for product improvement or development, because logging is only available in billing-enabled projects, which fall under the Paid Services terms. The Gemini API Additional Terms of Service state that, for Paid Services, Google does not use prompts, system instructions, cached content, files or responses to improve its products.

That protection changes when a developer chooses to share data. According to the same policy, datasets shared with Google are processed under the Unpaid Services terms and may be used to develop and improve Google products, services and machine learning, including training and evaluating future models. Human reviewers may read, annotate and process shared inputs and outputs, after Google disconnects the data from the Google Account, API key and Cloud project. Developers can opt out of sharing when creating a dataset, and Google asks them not to contribute logs that contain sensitive, confidential or proprietary information.

........

Data held by the Gemini API	Retention stated by Google	Developer control
Gemini API logs (paid tier)	55 days by default	7, 14, 28 or 55 days, or off
Logs saved to a dataset	No expiry	Delete or do not save
Stored interactions, free tier	1 day	store=false
Grounding with Google Search	30 days	Cannot be disabled
Grounding with Google Maps	30 days	Cannot be disabled
Live API session resumption	Up to 24 hours	Skip SessionResumptionConfig
File API uploads	Until deleted or expired	Delete manually

........

The policy also describes a Contributor Tier, an invite-only private preview in which eligible developers receive up to 50% off token costs on selected models, including Gemini 4 Argon and Gemini 3.8 Flash, in exchange for allowing data from those models in the project to be used under the Unpaid Services terms. This arrangement is separate from the new default, but it shows that Google treats developer traffic as valuable training material when developers explicitly agree to provide it.

··········

IMPLICATIONS FOR ZERO DATA RETENTION AND COMPLIANCE.

The new default has a direct bearing on teams that aim for zero data retention on the Gemini Developer API. Google's ZDR documentation describes zero data retention as a set of conditions rather than a single switch: customers must take specific actions or avoid specific features. It already lists the Interactions API, whose state storage is on by default and must be turned off with `store=false`. With the October 26 change, new paid projects using the Generate Content API need the same attention, either through the project setting or through the per-request parameter.

Even with logging off, the Gemini Developer API does not reach zero retention on its own. Google's terms state that, for Paid Services, prompts and responses are logged for a limited period solely to detect violations of its Prohibited Use Policy, and the documentation does not specify how long that period lasts. For guaranteed zero data retention or enterprise data processing agreements, Google's ZDR page directs customers to Vertex AI rather than to the Gemini Developer API.

There is also a documentation gap worth noting. Google's data logging and sharing policy, last updated on October 14, 2026, still describes logging as opt-in and states that the project owner chooses whether to enable it, and the logs documentation, last updated on September 23, 2026, still says the Generate Content API does not store requests by default. Both statements now describe only existing projects. Until those pages are revised, the release note is the authoritative description of the new default for paid projects created after the change.

··········

WHAT IS NOT LOGGED.

The logging feature has explicit exclusions, which means the new default does not apply uniformly to every workload a project may run. Google's documentation states that logging is not currently supported for Imagen and Veo models, Gemini embedding models, the Gemini Robotics model, inputs containing videos, GIFs or PDFs, and Public Preview Agents in the Gemini API. Managed Agents calls through the Interactions API are also excluded from the list of supported calls.

........

Workload	Logged under Gemini API logs
generateContent, streamGenerateContent, batchGenerateContent	Yes
Interactions API (except Managed Agents)	Yes
OpenAI compatibility endpoints	Yes
Imagen and Veo	No
Gemini embedding models	No
Gemini Robotics model	No
Inputs with videos, GIFs or PDFs	No
Public Preview Agents	No

........

These exclusions concern the developer-facing logs feature only. They do not change the abuse-monitoring logging or the grounding retention rules described in the Gemini API terms, which apply independently of the AI Studio Logs dashboard.

··········

TECHNICAL SYNTHESIS.

The verified facts are limited but clear. As of October 26, 2026, new paid-tier Gemini API projects store Generate Content API requests by default, matching the Interactions API; existing projects keep their current setting; and storage can be disabled per request with `store=false` or for the whole project in the Logs and Datasets settings. Google's documentation sets a 55-day default retention window that can be shortened, states that paid-tier logs are not used for product improvement unless the developer shares them in a dataset, and lists the models and inputs that logging does not cover.

Several points remain undisclosed. The release note does not explain why Google changed the default, does not define the exact cutoff for new projects, and does not say whether the change will later extend to existing projects. The duration of abuse-monitoring logs and the time between a log being marked for deletion and its removal are also not stated in the pages reviewed. For developers, the operational conclusion is narrower than the policy questions: any Gemini API project created from now on should have its logging setting reviewed before it handles user data, and applications that rely on `store=false` in the Interactions API should apply the same discipline to Generate Content calls.

··········

FOLLOW US FOR MORE.

·····

DATA STUDIOS

·····

datastudios.org