Skip to content
Free LLM Today

Do free LLM APIs train on your data?

What each provider’s own page says about using the prompts and answers of the free tier to train or improve models, quoted and dated.

Confirmed: does not train on your content (1)

Free LLM APIs that state they do not train on your content
ProviderTrains on your dataFree allowanceCommercial use
Cloudflare Workers AIFree tier
No

“Cloudflare does not use your Customer Content to train any AI models made available on Workers AI.”

10,000 Neurons/day

Resets at 00:00 UTC. Going above it needs Workers Paid ($0.011 per 1,000 Neurons).

Not verified

8 of 12 cells above were read on the provider’s own page in the last 30 days (oldest check: ). Each check mark opens that page. The other 4 say “not verified” instead of repeating a number from someone else’s article.

Confirmed: trains, or depends on your settings (2)

Confirmed: trains, or depends on your settings
ProviderTrains on your dataFree allowanceCommercial use
Google Gemini APIFree tier
Yes

On unpaid services, content is used “to provide, improve, and develop Google products”, and human reviewers may read it.

Not published

Limits are per project and shown only inside Google AI Studio.

Restricted

Only Paid Services may be used for API clients offered to users in the EEA, Switzerland or the UK.

OpenRouterFree tier
Your setting

An account setting decides whether requests may be routed to providers that train on your data, with a separate switch for free models.

20 req/min · 50 req/day

1,000 req/day once the account has bought at least 10 credits.

Not verified

No statement found (4)

The pages we read for these providers make no statement about training on free-tier content. Groq’s data page, for example, covers retention but not training.

What to check before you rely on it

Training, retention and review are three questions

Training is whether your content improves a model. Retention is how long it is stored. Review is whether a person may read it. Google’s terms for unpaid Gemini API services answer yes to training and to review. Groq’s data page answers the retention question (by default it does not retain customer data for inference requests, and Zero Data Retention can be enabled) and is silent on training.

On a router, the answer is the upstream provider’s

OpenRouter and Hugging Face pass your request to another company. OpenRouter lets you forbid routing to providers that may train on your data, with a separate switch for free models. With that switch off, fewer free endpoints may be available to you.

Paying usually changes the answer

On the Gemini API, content sent on the paid tier is not used to improve Google products. If the data matters, the price of privacy is the paid tier, not a different free one. For what you may build on each tier, see commercial use.

Questions and answers

Do free LLM APIs train on my prompts?

Some do. Google’s Gemini API terms say content sent to unpaid services is used to provide, improve and develop Google products, and that human reviewers may read it. Cloudflare states it does not use Customer Content to train models on Workers AI. On OpenRouter it depends on an account setting.

How do I keep prompts out of training on OpenRouter?

In the account privacy settings you choose whether requests may be routed to providers that may train on your data, with a separate switch for free models. With training disallowed, OpenRouter does not route to providers that train.

What does Groq do with my data?

Its data page says that, by default, Groq does not retain customer data for inference requests, and that Zero Data Retention can be enabled in Data Controls. The page does not make a statement about training, so that cell says “not published”.

Should I send personal or confidential data to a free tier?

Not where the terms allow training or human review, as with the Gemini API free tier. Use a paid tier with a no-training commitment, or a provider whose page states it does not train on your content.

Related pages