CloudLast verified September 26, 2026

Groq data retention and zero data retention (ZDR)

Groq does not keep your inputs or outputs by default, and does not train on them on any tier. It may log requests for up to 30 days when troubleshooting or investigating abuse. You can turn that logging off yourself in Data Controls, which is zero data retention (ZDR).

Quick policy snapshot

Default retention: See vendor policy
Zero data retention available: Yes
Trains on API customer data by default: No

Groq's training default

Groq's Services Agreement is explicit: Groq is not permitted to use customer Inputs or Outputs to train or fine-tune any AI model unless you explicitly grant permission or instruct it to. Groq is an inference provider, not a foundation-model developer, so the company does not run a research-data pipeline that would consume your prompts. This posture is account-wide and is not split between free and paid tiers, so the no-training default applies to your BYOK traffic regardless of plan.

When Meetily users select a Groq-hosted model as their summary provider via BYOK, the request hits api.groq.com under your API key. The defaults on this page apply.

Retention

By default, Groq does not retain inference Inputs and Outputs. It may temporarily log them only when troubleshooting errors that degrade platform reliability or when investigating suspected abuse, and those logs are kept for up to 30 days unless law requires longer. You can opt out of that logging entirely in the Data Controls settings.

Application-state features carry their own windows: batch processing files are retained for 30 days unless deleted earlier, and fine-tuning models and datasets are kept until you delete them. After contract termination, Groq's DPA commits to deleting Personal Data on request within a maximum of 180 days. Review the current agreement before deployment in regulated environments.

Zero data retention

Zero Data Retention is now self-serve for every customer. An organization admin can enable ZDR in the Data Controls settings of the Groq Console, either globally or per feature, at any time. When ZDR is on, Groq does not retain customer data for system reliability or abuse monitoring. This is a change from the older posture, where zero-retention terms were available only by negotiating through enterprise sales.

How Meetily uses Groq

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

Groq is a reasonable cloud-summary target for Meetily users who want a low-latency BYOK provider that does not train on their data, and you can turn on Zero Data Retention yourself in the Console before sending transcripts. For zero retention by construction, switch to local Ollama and keep the entire pipeline on-device.

References

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

"Your Data in GroqCloud", Groq (GroqDocs). https://console.groq.com/docs/your-data - accessed 26 September 2026. Source for default no-retention of inference data, the up-to-30-day troubleshooting and abuse log window, batch and fine-tuning retention, and self-serve Zero Data Retention in Data Controls.
"Data Processing Addendum for GroqCloud Services", Groq (GroqDocs). https://console.groq.com/docs/legal/customer-data-processing-addendum - accessed 29 June 2026. Source for the purpose-limited processing terms and the 180-day post-termination deletion commitment.
"Groq Trust Center", Groq. https://trust.groq.com - accessed 29 June 2026. Source for Groq's compliance posture and security controls.
"Privacy Policy", Groq. https://groq.com/privacy-policy - accessed 29 June 2026. Corroborating source for Groq's general data handling and the no-sell, no-share commitments.
"Groq Services Agreement", Groq (GroqDocs). https://console.groq.com/docs/legal/services-agreement - accessed 26 September 2026. Source for the commitment that Groq is not permitted to use Inputs or Outputs to train or fine-tune any model unless the customer grants permission or instructs it to.

Last verified: September 26, 2026. Policy source: Groq Cloud policy

Frequently asked questions

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

No, on either tier. Groq's Services Agreement states it is not permitted to use customer Inputs or Outputs to train or fine-tune any AI model unless you explicitly grant permission or instruct it to. Groq is an inference provider rather than a foundation-model developer, so the same no-training posture applies to free developer usage and paid usage; the policy is account-wide, not tier-gated. The defaults on this page apply to your BYOK traffic.

What is the default retention period for Groq API traffic?

By default Groq does not retain inference Inputs and Outputs. It may temporarily log them when troubleshooting reliability issues or investigating suspected abuse, and those logs are kept for up to 30 days unless law requires longer; you can opt out of that logging in Data Controls. Batch processing files are retained 30 days unless deleted earlier, and fine-tuning models and datasets are kept until you delete them. After contract termination, the DPA commits to deletion within a maximum of 180 days.

Is zero data retention available for Groq?

Yes, and it is now self-serve. Any customer can enable Zero Data Retention in the Data Controls settings of the Groq Console, globally or per feature. When ZDR is on, Groq does not retain customer data for system reliability or abuse monitoring. This is a change from the older posture where ZDR-style terms were negotiated only through enterprise sales.

Where is Groq API data stored geographically?

Groq's DPA states data may be transferred and processed in the United States and other countries where Groq operates. Groq's primary infrastructure is US-based. Confirm current region options in the Trust Center at trust.groq.com.

How do I request deletion of data Groq has retained?

Manage retention directly in the Data Controls settings of the Groq Console, including enabling Zero Data Retention and disabling batch processing or fine-tuning. Upon termination of service, Groq's DPA commits to deleting Personal Data on request within a maximum of 180 days unless legal obligations require longer retention.

Does Groq have SOC 2 or other compliance certifications?

Groq publishes its compliance posture on their Trust Center at trust.groq.com. SOC 2 Type 2 is the most relevant baseline for cloud LLM inference providers; check the current certifications list before regulated deployments.

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

Meetily transcription runs 100% locally by default. When you select a Groq-hosted model as your summary provider via BYOK, Meetily routes transcript text (not audio) to api.groq.com using your own API key. The retention and training defaults on this page apply to your traffic.

Why is Groq popular for low-latency summarization?

Groq's LPU inference hardware delivers very low time-to-first-token on supported open-weight models (Llama, Mixtral, Gemma). For Meetily users who want a fast cloud summary path but still need an API that does not train on their data, Groq is a reasonable BYOK target. For zero retention by construction, switch to local Ollama instead.

Use Groq Cloud with Meetily, on your terms

Meetily transcription stays 100% local by default. For summaries, bring your own Groq Cloud key (BYOK) so the data path matches the policy you just read - or pick a local model if you want zero retention by construction.

Set up BYOK Download Meetily