Gemini API now stores prompts by default on new paid projects
Since October 26, new paid Gemini API projects log generateContent requests by default for 55 days. How to set store=false and keep customer data out.
Google changed a default in the Gemini API: since October 26, new paid-tier projects store generateContent requests by default. That means full prompts and responses sit in AI Studio logs for up to 55 days unless you turn it off. If your app sends customer names, orders or support tickets to Gemini, a new project now keeps a copy of them.
What actually happened
Per the Gemini API release notes, generateContent, streamGenerateContent and batchGenerateContent now default to store=true on new paid projects. That matches the Interactions API, which already stored by default. Projects created before October 26 keep their current setting, which is off unless you turned logging on yourself.
The logs and datasets guide fills in the rest:
- What's stored: the full prompt, the full response and the context from earlier turns.
- Retention: 55 days by default. You can set a project to 7, 14, 28 or 55 days. Logs you save to a dataset don't expire.
- Opt out per project: in AI Studio's Logs settings, toggle Generate Content and Interactions logging separately.
- Opt out per request: pass
store=false. For Generate Content that needs a recent SDK: google-genai 2.33.0+ (Python), @google/genai 2.31.0+ (JavaScript), 1.79.0+ (Java), or 1.76.0+ (Go). - Catch: turning Interactions API logging off also stops automatic conversation history, unless a request overrides it.
The same release notes list a Gemini API CLI (gemini-api) beta on October 27, built for coding agents and CI pipelines. If an agent spins up a fresh project for it, that project gets the new default.
Why it matters for your business
Logging is useful. You can debug a bad answer, build an eval set and see what users actually ask. But it's also a second copy of your customer data, held by a vendor, for nearly two months. That copy belongs in your privacy policy and your data map, and you should decide on it deliberately.
What we'd do this week:
- List your Gemini projects and note which ones were created on or after October 26.
- Pick a policy per app. Internal tools with no personal data: logging on, short window. Anything with customer PII:
store=falsein code, not just a dashboard toggle someone can flip back. - Pin your SDK version so the per-request flag actually works.
- Check your saved datasets. Those never expire.
Key takeaways
- New paid Gemini API projects store generateContent requests by default since October 26, 2026
- Logs keep full prompts and responses for 55 days by default; saved datasets never expire
- Older projects keep their existing setting
- Opt out per request with store=false (recent SDKs) or per project in AI Studio
- Set the policy in code for any app that sends customer data
Know where your AI vendors keep your customers' data? We build AI integrations with retention set in code and written down, so nothing changes because a vendor flipped a default. See how we build, or send us your stack for a quick data-flow check.
Sources: Gemini API release notes, Gemini API logs and datasets guide.
- #gemini-api
- #data-retention
- #privacy
- #ai-logging
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
AKV cache engineering: cut AI agent costs 27.7% vs summarizing
Tsinghua's open-source AKV lets AI agents drop old tool outputs from the KV cache instead of summarizing. Authors report 27.7% lower cost. What to copy now.
Read itOpera WebAI: local AI in the browser, free inference per call
Opera WebAI ships local Gemma models in Opera One with Prompt, Decision and Embeddings APIs. Zero per-request cost, opt-in only. Build a server fallback.
Read it