KoolitusAI ✳︎
← All posts

Tools · Published 2 November 2026 · Gemini API · Andmekaitse · Google

New Gemini API projects now keep every prompt and response for 55 days by default

This post was written by AI from public sources. The sources are listed at the end.

Since 26 October, paid Gemini API projects created on or after that date store full requests and answers in Google AI Studio unless you switch it off. If your app sends customer data to Gemini, check your settings this week.

Google changed a default in the Gemini API on 26 October 2026. Paid-tier projects created on or after that date now store Generate Content API requests by default. Projects created earlier keep their old setting, which was off unless someone turned logging on. The Interactions API already stored requests by default.

A stored request is more than a line in a usage report. According to Google's documentation, the Logs page in AI Studio shows the full prompt, the model's answer and the context from earlier turns. Calls made through the OpenAI compatibility endpoints are logged too. Inputs containing video, GIFs or PDFs are not logged, and neither are Imagen, Veo or the embedding models.

How long the data stays, and who else uses it

Logs expire after 55 days by default. You can shorten that to 7, 14 or 28 days per project. Logs you save into a dataset never expire. Google's logs policy says that for billing-enabled projects, logged prompts and responses are not used for product improvement. That changes if you share a dataset or feedback with Google: shared data falls under the terms for unpaid services, may be used to improve Google's models, and human reviewers may read it after it is separated from your account.

The logs policy does not say where the logs are stored. The general Gemini API terms say logged data may be stored or cached in any country where Google or its agents have facilities, and that apps offered to users in the European Economic Area may only use the paid services.

Why this matters for an Estonian team

Many internal tools built this autumn send real material to Gemini: customer emails, support tickets, contracts, CVs. A developer who created a new project after 26 October and changed nothing now has that material sitting in a log for almost two months. The logs can help with debugging and quality checks. They also count as personal data your company holds, and your data protection records probably do not mention them.

You can switch storage off for the whole project in the Settings panel of the AI Studio Logs page, or per request by setting store to false. The request setting wins over the project setting. On Generate Content calls it needs a recent SDK: google-genai 2.33.0 or later for Python, @google/genai 2.31.0 or later for JavaScript. Turning off logging for the Interactions API also stops it from keeping conversation history for you, so test before you change it in production.

What to do on Monday

  • List every Gemini API project your company uses and note which ones were created on or after 26 October 2026
  • Open the Logs page in AI Studio for each project and see what is already stored
  • Decide per project whether you need the logs. If you keep them, cut retention to 7 or 14 days
  • For requests that carry personal or confidential data, update the SDK and send store set to false
  • Check who has access to those Cloud projects, because the logs show full prompts and answers
  • Do not share datasets or feedback with Google from projects that handle customer data
  • Add Gemini logs to your record of processing activities, with the retention period you chose

The change took effect quietly, through a release note. Your developers can fix it in an afternoon, but someone has to ask them first.

Key takeaways

  • Paid Gemini API projects created on or after 26 October 2026 store full prompts, answers and conversation context by default
  • Logs expire after 55 days unless you shorten the window; data you share with Google can be used to improve its models
  • Check your projects, shorten retention or set store to false for personal data, and add the logs to your data protection records

Related training

AI Agents and Workflow Automation

Build AI agents and automations on your own processes, tested and running by the end of day two.

Book this training

Loe eesti keeles