Gemini Zero Data Retention: Paid API vs Free Tier

Gemini API zero data retention (ZDR) and training

The paid Gemini API and Vertex AI do not train on your prompts; the free Google AI Studio tier can, outside the EEA, Switzerland and the UK. Even on paid tiers, Google logs prompts for a limited time to detect abuse. For guaranteed zero data retention (ZDR), Google points you to Vertex AI. This page covers what is stored, the 24-hour cache, and the steps to reach ZDR.

Gemini's two-tier privacy model

Google operates Gemini in two distinct privacy regimes:

Paid Gemini API and Vertex AI: Google does not use prompts or responses to improve their products. Data is processed under the Google Cloud Data Processing Addendum.
Free Google AI Studio tier: Google uses submitted content to improve products and services, including for machine-learning purposes, and reviewers may annotate inputs and outputs. The terms explicitly warn against submitting sensitive, confidential, or personal data on this tier.

Customers in the European Economic Area, Switzerland, and the UK get the paid-tier policy applied to all services including free tiers, per Google's regional commitments.

When Meetily users select Gemini as their summary provider via BYOK, the tier you are on is determined by your API key. Use a Vertex AI or paid Gemini API key for production traffic.

Retention

On the paid Gemini API and Vertex AI, prompts and responses are logged for a limited period solely for safety, abuse detection, and legal compliance. Google's terms note the data may be stored transiently or cached in any country where Google maintains facilities, subject to your project's region configuration.

Specific features have their own retention windows. On the Gemini Developer API, grounding with Google Search or Google Maps keeps data for 30 days, the Interactions API stores state unless you set store to false, and uploaded files stay until you delete them. On Vertex AI, grounding with Google Search keeps query logs for up to 3 days for debugging, and that storage cannot be disabled.

Both surfaces also cache inputs and outputs in memory for up to 24 hours to speed up responses. The cache is in RAM only, isolated to your project, and Google states it does not break zero data retention. On Vertex AI you can switch it off at the project level.

Zero data retention

Zero data retention on Gemini is something you configure, not a single switch. Google's documentation now also uses the name Gemini Enterprise Agent Platform for Vertex AI.

On Vertex AI, reaching ZDR means:

Requesting an exception to prompt logging for abuse monitoring.
Not enabling Live API session resumption, which stores session data for up to 24 hours.
Avoiding grounding with Google Search, whose query logs (up to 3 days) cannot be turned off. Google recommends Web Grounding for Enterprise instead.
Checking with your account team if you use advanced models, which may carry extra logging.

On the paid Gemini Developer API, Google logs prompts and responses for a limited time to detect abuse, and its ZDR guide says to use Vertex AI if your workload requires guaranteed zero data retention. To get as close as possible on the Developer API, set store to false on the Interactions API, avoid context caching and session resumption, and delete uploaded files yourself.

Region pinning is configured at Vertex AI project creation and is the primary lever for data residency.

How Meetily uses Gemini

Meetily routes Gemini traffic through your own API key. Your project's tier and contract terms apply. The transcript text is sent over TLS to the Gemini endpoint for summarization, and the response is returned to Meetily and stored locally on your device. Audio is never transmitted to Google at any point.

If your project is on the free Google AI Studio tier, move to paid Gemini API or Vertex AI before sending any meeting transcripts that contain sensitive content. For stronger guarantees, switch your Meetily summary model to a local provider (Ollama) and keep the entire pipeline on-device.

References

This page was last checked on 26 September 2026 against sources 1, 5 and 6, and on 29 June 2026 against the others. Always confirm the current terms against Google's own linked policy before a regulated deployment.

"Gemini API Additional Terms of Service", Google (Google AI for Developers). https://ai.google.dev/gemini-api/terms - accessed 26 September 2026 (last modified 28 April 2026). Source for the paid-vs-unpaid data-use split and the EEA / Switzerland / UK exception.
"Gemini Developer API pricing", Google. https://ai.google.dev/gemini-api/docs/pricing - accessed 29 June 2026. Distinguishes the free and paid API tiers referenced in the training split.
"Google Cloud compliance offerings", Google Cloud. https://cloud.google.com/security/compliance - accessed 29 June 2026. Source for the Vertex AI / Google Cloud compliance posture (SOC 2, HIPAA BAA eligibility, ISO 27001).
"Google Cloud Platform Data Processing Addendum (Customers)", Google Cloud. https://cloud.google.com/terms/data-processing-addendum - accessed 29 June 2026. Source for Vertex AI data residency, deletion procedures, and ZDR-equivalent contractual terms.
"Zero data retention in the Gemini Developer API", Google (Google AI for Developers). https://ai.google.dev/gemini-api/docs/zdr - accessed 26 September 2026 (last updated 14 September 2026). Source for abuse logging on paid services, the "use Vertex AI" guidance for guaranteed ZDR, the Interactions API store setting, the 24-hour in-memory cache, and 30-day grounding retention.
"Gemini Enterprise Agent Platform and zero data retention", Google Cloud. https://docs.cloud.google.com/vertex-ai/generative-ai/docs/vertex-ai-zero-data-retention - accessed 26 September 2026 (last updated 25 September 2026). Source for the abuse-monitoring exception, session resumption, the 3-day Search-grounding logs, advanced-model logging, and project-level cache control on Vertex AI.

Frequently asked questions

Does Google train Gemini models on my API prompts? (free vs paid)

It depends on the tier, and the split is the single most important thing to get right. On the PAID Gemini API and Vertex AI, Google's terms state Google does not use your prompts or responses to improve its products, which includes training. On the FREE Google AI Studio tier and unpaid Gemini API quota, Google does use submitted content to provide, improve, and develop Google products and services, and human reviewers may see it. One exception: if you are in the European Economic Area, Switzerland, or the UK, the paid-services data terms apply to all services including the free tiers. To be safe with meeting transcripts, use a paid Gemini API or Vertex AI key in Meetily.

What is the default retention period for Gemini API traffic?

On the paid Gemini API and Vertex AI, prompts and responses are logged for a limited period solely for detecting abuse and policy violations, and for legal disclosures. Google does not publish a fixed number of days for that log. Some features keep data longer: on the Gemini Developer API, grounding with Google Search or Google Maps keeps data for 30 days; on Vertex AI, grounding with Google Search keeps query logs for up to 3 days, and that storage cannot be turned off.

Is zero data retention available for Gemini?

Yes, on Vertex AI, if you take specific steps. You can request an exception to prompt logging for abuse monitoring, you should not enable Live API session resumption, and you should avoid grounding with Google Search, which keeps query logs for up to 3 days with no way to switch that off. Some advanced models may add their own logging, so check with your Google Cloud account team. On the paid Gemini Developer API, Google still logs prompts for abuse detection, and its ZDR guide says to use Vertex AI if your workload requires guaranteed zero data retention.

Where is Gemini API data stored geographically?

Vertex AI offers region selection at project creation; data residency commitments apply per the Google Cloud DPA. The free Google AI Studio tier processes data on Google global infrastructure without region pinning.

How do I request deletion of data Google has retained?

Use Google Cloud's data deletion procedures documented in the Google Cloud DPA. For Google AI Studio free-tier data, use the privacy controls in your Google Account.

Does Google Cloud / Vertex AI have HIPAA or SOC 2 compliance?

Yes. Google Cloud publishes its compliance posture (SOC 2 Type 2, HIPAA BAA eligibility for covered services, ISO 27001, and more) on cloud.google.com/compliance. Vertex AI Gemini endpoints fall under the Google Cloud compliance umbrella.

How does Meetily handle Gemini when I pick it as my summary provider?

Meetily transcription runs 100% locally by default. When you select Gemini as your summary provider via BYOK, Meetily routes transcript text (not audio) to the Gemini endpoint using your own API key. The training and retention defaults on this page apply to your traffic.

What if I want zero retention but my project is on the free Google AI Studio tier?

Move to Vertex AI (paid) and use a project with ZDR-equivalent terms in the DPA, or switch your Meetily summary model to a local provider like Ollama. The local-summary path keeps the entire pipeline on-device.