What changed
On October 26, 2026, Google changed the default for new paid-tier Gemini API projects using Generate Content methods: generateContent, streamGenerateContent and batchGenerateContent now store request logs (`store=true`) unless explicitly disabled. The Interactions API already stored requests by default. Projects created before October 26 retain their existing Generate Content setting, which is off unless the owner previously enabled it. Google's logs documentation confirms that developers can inspect full prompts, responses and previous-turn context in AI Studio; logs expire after a default 55-day retention period, with 7-, 14-, 28- and 55-day options. Logs saved to datasets do not expire automatically.
Why it matters
A new production project can retain user prompts, customer documents and generated responses even when its predecessor did not. That affects data mapping, privacy disclosures, vendor reviews, access controls and internal incident-response assumptions. It is not evidence that Google uses those logs to train models: Google's documentation treats dataset sharing as an optional separate step. The change is actionable because developers can set store=false per request or turn off Generate Content logging for the project. Teams should verify effective settings, not assume older projects were migrated.
The change applies to new paid projects, not all existing API users
Google's October 26 changelog explicitly scopes the new default to new paid-tier projects calling generateContent, streamGenerateContent and batchGenerateContent. Projects created before that date keep their Generate Content setting; it remains off unless previously enabled. The Interactions API has a separate logging toggle and already defaults to storage because it also supports server-managed conversation state. Google says log storage is available only on the Gemini API paid tier.
What the logs contain and how long they last
The AI Studio Logs dashboard can display a request's full prompt, model response and prior-turn context. Logs are marked for deletion after 55 days by default, or an administrator can select 7, 14, 28 or 55 days. Developers can save selected logs into datasets for evaluation or export; those datasets do not automatically expire with the log window. Logging currently excludes some media and models, including Imagen/Veo, embeddings, robotics, inputs with videos/GIFs/PDFs and preview agents.
Two independent controls for opt-out
A request-level store=false overrides project settings. Google documents support for this parameter in Python google-genai 2.33.0+, JavaScript @google/genai 2.31.0+, Java google-genai 1.79.0+ and Go google.golang.org/genai 1.76.0+. Alternatively, a project owner can disable Generate Content API logging from the Settings panel of the Logs and Datasets dashboard. The Generate Content and Interactions API toggles are independent; disabling Interactions storage affects its server-side conversation history feature unless overridden per request.
Audit privacy and observability together
Inventory projects created since October 26, check which SDK versions and API methods they use, and verify the effective logging setting in AI Studio. Confirm who can view and export logs or save non-expiring datasets. If sensitive user content should not be retained in these logs, disable storage and test the setting in a controlled request. If logs are intentionally retained for debugging, document the selected retention period, project access and dataset governance. This change is about request storage in AI Studio, not a documented alteration to Google's model-training policy.