---
title: "GDPR LLM API checklist - 15 questions for providers | BroadRouter"
description: "A GDPR checklist for LLM APIs: DPA terms, subprocessors, retention, training, international transfers and breach notice, with the GDPR articles behind each question."
url: https://broadrouter.com/blog/gdpr-llm-api-checklist/
lang: en
updated: 2026-11-14
publisher: BroadRouter (a BroadNet site); API by BroadNet Inference, https://inference.broadnet.ai
---

# GDPR and LLM APIs: 15 questions to ask your provider

By the BroadNet Inference team · 5 min read · Published 2026-11-14 · Updated 2026-11-16

Before you send personal data to an LLM API, get a signed data processing agreement (GDPR Article 28), the subprocessor list, written retention and training terms, the processing location and the legal basis for any transfer outside the EEA (Articles 44 to 49). Then minimise what you send. This is a checklist, not legal advice.

## Key takeaways

-   An LLM API that processes personal data on your behalf is normally your processor, so Article 28 requires a written contract.
-   Ask for the subprocessor list and how you are told about changes. GPU hosts and backup capacity count.
-   Retention and training are separate questions. Get both answers in writing.
-   Transfers to the US need a basis: the EU-US Data Privacy Framework for listed companies, or Standard Contractual Clauses plus a transfer assessment.
-   The cheapest GDPR control is data minimisation: do not send personal data the model does not need.

**On this page**

1.  Is an LLM API provider a processor under GDPR?
2.  What 15 questions should you ask an LLM provider under GDPR?
3.  Why are retention and training separate questions?
4.  International transfers: where do prompts go?
5.  How do you minimise personal data before sending it?
6.  What are the red flags in a provider’s GDPR answers?
7.  Where BroadNet Inference fits

Before personal data goes to an LLM API, you need a signed data processing agreement (Article 28), the provider’s subprocessor list, written answers on retention and training, the processing location and a transfer basis for anywhere outside the EEA (Articles 44 to 49). Then send as little personal data as possible. You are usually the **controller**, the provider your **processor**.

This checklist turns the relevant GDPR articles into questions you can send to any provider. It is not legal advice. Your data protection officer or counsel decides what is enough for your case.

## Is an LLM API provider a processor under GDPR?

In the common case, yes. You decide why the data is processed (to summarise a ticket, to answer a customer), and the provider processes it for you. Article 28 then requires a written contract with specific terms. If a provider also uses your prompts for its own purposes, such as training its models, it may be acting as a controller for that part. That is one reason the training question matters so much.

The EDPB’s [Opinion 28/2024](https://edpb.europa.eu/system/files/2024-12/edpb_opinion_202428_ai-models_en.pdf) deals with personal data in AI models themselves. The [LLM risk report](https://www.edpb.europa.eu/our-work-tools/our-documents/support-pool-experts-projects/ai-privacy-risks-mitigations-large_en) from the EDPB’s Support Pool of Experts (10 April 2025) works through privacy risks and mitigations for LLM systems, with three use cases: a customer-query chatbot, a student-progress monitor and a travel assistant. It was written by external experts and states that it does not necessarily reflect the EDPB’s official position, but it is a practical starting point.

## What 15 questions should you ask an LLM provider under GDPR?

| # | Ask the provider | Why (GDPR) |
| --- | --- | --- |
| 1 | Will you sign a DPA, and is it part of the terms or a separate document? | Art. 28(3) |
| 2 | Do you process personal data only on our documented instructions? | Art. 28(3)(a) |
| 3 | Who are your subprocessors, including GPU hosts and backup capacity? | Art. 28(2), 28(4) |
| 4 | How are we told about new subprocessors, and can we object? | Art. 28(2) |
| 5 | Are prompts, files, audio or outputs stored after the request? Where and for how long? | Art. 5(1)(e) |
| 6 | Is any of our content used to train, fine-tune or evaluate models? Is that off by default? | Art. 5(1)(b), Art. 28(10) |
| 7 | Which records do you keep without content (IP, token counts, cost), and for how long? | Art. 5(1)(c) |
| 8 | In which countries is data processed, including failover? | Art. 44 |
| 9 | What is the transfer basis: adequacy, DPF certification or SCCs? | Art. 45, 46 |
| 10 | Can we pin processing to a region? | Art. 44 |
| 11 | What security measures apply to requests and keys? | Art. 32 |
| 12 | How fast will you tell us about a personal data breach? | Art. 33(2) |
| 13 | How do you help with access and erasure requests? | Art. 28(3)(e) |
| 14 | Do staff ever see request content, for example on a support ticket? | Art. 28(3)(b), 32 |
| 15 | Will you give us the information we need for a DPIA and allow audits? | Art. 28(3)(f), 28(3)(h), Art. 35 |

Ask for answers in writing and keep them with your records of processing (Article 30).

## Why are retention and training separate questions?

Providers often answer one and imply the other. Get both:

-   **Retention:** is content stored after the request, in logs, abuse-monitoring stores, caches, batch files or conversation state?
-   **Training:** is content used to improve models, and is the default on or off for your account type?

Policies differ a lot between providers and between free and paid tiers. We compared the official wording of five providers in [LLM API data retention compared](https://broadrouter.com/blog/llm-api-data-retention-compared/). A [zero data retention](https://broadrouter.com/glossary/zero-data-retention/) arrangement answers the retention question well, but you still need a contract and a transfer basis. Our [ZDR checklist](https://broadrouter.com/blog/zero-data-retention-llm-api-checklist/) covers the technical side.

## International transfers: where do prompts go?

If the provider processes prompts outside the EEA, Chapter V of the GDPR applies. In practice:

1.  **Adequacy.** The Commission keeps a [list of adequacy decisions](https://commission.europa.eu/law/law-topic/data-protection/international-dimension-data-protection/adequacy-decisions_en). For the US, the EU-US Data Privacy Framework decision (10 July 2023) covers only companies on the [Data Privacy Framework List](https://www.dataprivacyframework.gov/list). Check the exact legal entity you contract with.
2.  **Standard Contractual Clauses.** If the company is not on the list, the usual basis is the Commission’s [SCCs](https://commission.europa.eu/law/law-topic/data-protection/international-dimension-data-protection/standard-contractual-clauses-scc_en), with a transfer impact assessment.
3.  **Onward transfers.** Ask where subprocessors process data. A provider in one country can fail over to capacity in another.

Other regimes work differently. Egypt, for example, requires a licence or permit from its data protection regulator for most transfers abroad; see [Egypt’s data protection law and LLM APIs](https://broadrouter.com/blog/egypt-data-protection-law-llm-api/). For Saudi Arabia and the UAE, see the [PDPL checklist](https://broadrouter.com/blog/saudi-uae-pdpl-llm-api-checklist/).

## How do you minimise personal data before sending it?

The strongest control is the one you run yourself. Article 5(1)(c) requires data minimisation, and it also shrinks your exposure to every answer above:

-   Strip names, emails, phone numbers and IDs the model does not need.
-   Replace them with placeholders and map them back after the response.
-   Never send special-category data (Article 9) unless you have a basis and a clear need.
-   Keep your own logs free of prompt content.

We cover redaction, placeholders and logging in detail in [sending sensitive data to an LLM API](https://broadrouter.com/blog/llm-api-sensitive-data/).

## What are the red flags in a provider’s GDPR answers?

-   “We are GDPR compliant” with no DPA to sign.
-   No subprocessor list, or one that leaves out the GPU host.
-   Training “may” happen, with no clear default for API customers.
-   Retention described only as “a limited period”.
-   No answer on where failover traffic runs.

## Where BroadNet Inference fits

BroadNet’s [privacy statement](https://www.broadnet.ai/inference-privacy) (effective September 2026) answers several of the questions directly. It says the gateway “keeps no prompts, messages, images, audio, responses or transcripts”, and that “We do not use your inference data to train, fine-tune or evaluate models.” It keeps content-free usage records “for as long as billing, tax, security and legal obligations require”. Details are on our [zero data retention page](https://broadrouter.com/zero-data-retention/).

On location, the statement says BroadNet’s own and hosted open-weight models run “on BroadNet-owned or leased hardware in the United States, with partner capacity as backup”, and that a current subprocessor list “is available to customers on request”. For residency needs it asks customers to agree processing locations before sending data.

What it does not cover: the privacy statement does not mention a DPA or the GDPR by name. If you process EU personal data, ask BroadNet for its DPA, its subprocessor list and its transfer basis before you go live, exactly as you would with any provider on this list. You can also lock keys to your servers with [IP allow-lists](https://broadrouter.com/guides/api-keys-ip-allow-lists/). For the wider provider decision, see the [20-point LLM API checklist](https://broadrouter.com/blog/choose-llm-api-provider-checklist/).

Free tools for this

-   [OpenAI-compatible API request builder](https://broadrouter.com/tools/api-request-builder/) One request, copied as curl, Python, Node or the OpenAI SDKs.
-   [LLM API cost calculator](https://broadrouter.com/tools/llm-api-cost-calculator/) Price a workload on every model at once, with cache hits, reasoning and a flat-plan check.

[All free tools](https://broadrouter.com/tools/)

## Terms in this article

[GDPR](https://broadrouter.com/glossary/gdpr/)

The GDPR (General Data Protection Regulation, Regulation (EU) 2016/679) is the EU law on processing personal data. It has applied since 25 May 2018 and covers EU-established organisations and some outside the EU. For LLM APIs it governs processor contracts, international transfers and data minimisation.

[Subprocessor](https://broadrouter.com/glossary/subprocessor/)

A subprocessor is a third party that a processor engages to carry out processing of personal data on behalf of the controller. For an LLM API, subprocessors can include GPU hosts, partner inference providers and backup capacity. Under GDPR Article 28 they need the controller's authorisation and the same data protection obligations.

[Data processing agreement (DPA)](https://broadrouter.com/glossary/dpa/)

A data processing agreement (DPA) is the contract between a controller and a processor that governs how the processor handles personal data. GDPR Article 28(3) requires it and lists its contents: documented instructions, confidentiality, security, subprocessor rules, deletion or return of data, and audits.

[Zero data retention](https://broadrouter.com/glossary/zero-data-retention/)

Zero data retention (ZDR) is a data-handling policy in which an AI provider processes prompts and outputs only to serve a request and does not store that content afterward, in logs, caches or training data. Content-free records such as token counts and timings are usually still kept.

[Inference provider](https://broadrouter.com/glossary/inference-provider/)

An inference provider is a company that runs machine learning models, often open-weight LLMs, on GPUs it owns, leases or rents, and sells access to them through an API, usually priced per token, per request or per unit of time. It handles serving, scaling and hardware.

[Open-weight model](https://broadrouter.com/glossary/open-weight-model/)

An open-weight model is a machine learning model whose trained weights are published for anyone to download and run, usually under a licence with conditions. It differs from fully open-source AI, which also releases the training code and detailed information about the training data, so others can study and reproduce the model.

[PDPL (Saudi and UAE data protection laws)](https://broadrouter.com/glossary/pdpl/)

PDPL (Personal Data Protection Law) is the common name for Saudi Arabia's personal data law, issued by Royal Decree M/19 and enforced by SDAIA, and for the UAE's Federal Decree-Law No. 45 of 2021. Both regulate processing of personal data and set conditions for transfers abroad, including prompts sent to foreign LLM APIs.

[Token](https://broadrouter.com/glossary/token/)

A token is the unit of text a language model reads and writes: a word, part of a word, a punctuation mark or a byte sequence produced by the model's tokenizer. LLM APIs count, limit and usually bill usage in tokens.

## Questions

### Is using the OpenAI-compatible API of an LLM provider GDPR compliant?

No API is compliant or non-compliant by itself. Compliance depends on your legal basis, your contract with the provider, where data is processed, how long it is kept and what you send. Use the checklist in this post and involve your DPO.

### Do I need a DPA with an LLM API provider?

If the provider processes personal data on your behalf, yes. GDPR Article 28(3) requires a contract that binds the processor, including to act only on your documented instructions and to delete or return data at the end of the service.

### Is sending prompts to a US LLM API an international transfer?

If prompts contain personal data and the processing happens outside the EEA, yes. You need a Chapter V basis, such as the EU-US Data Privacy Framework for a certified company, or Standard Contractual Clauses with a transfer assessment.

### Does zero data retention make an LLM API GDPR compliant?

It helps with storage limitation and lowers breach risk, but the provider still processes personal data during the request. You still need a contract, a legal basis and a transfer mechanism.

### Does the EDPB have guidance on LLMs?

Yes. The EDPB adopted Opinion 28/2024 on personal data in AI models, and its Support Pool of Experts published a report on privacy risks and mitigations for LLMs in April 2025. The report states it is not an official EDPB position.

## Keep reading

-   [Saudi and UAE PDPL and LLM APIs: a cross-border checklist](https://broadrouter.com/blog/saudi-uae-pdpl-llm-api-checklist/)
-   [Egypt's data protection law (Law 151 of 2020) and LLM APIs](https://broadrouter.com/blog/egypt-data-protection-law-llm-api/)
-   [Zero data retention LLM API: a 10-question checklist](https://broadrouter.com/blog/zero-data-retention-llm-api-checklist/)
-   [LLM API data retention compared: OpenAI, Anthropic, Gemini, Mistral](https://broadrouter.com/blog/llm-api-data-retention-compared/)
-   [Sending sensitive data to an LLM API: redact, minimise, log less](https://broadrouter.com/blog/llm-api-sensitive-data/)
-   [How to choose an LLM API provider: a 20-point checklist](https://broadrouter.com/blog/choose-llm-api-provider-checklist/)
-   [How to test an LLM for sensitive-data safety: BroadNet's method](https://broadrouter.com/blog/llm-sensitive-data-safety-evaluation-method/)
-   [The Safety Map: how 39 LLMs handle sensitive data](https://broadrouter.com/blog/llm-safety-map-39-models/)
-   [Guide: Lock API keys to your servers with IP allow-lists](https://broadrouter.com/guides/api-keys-ip-allow-lists/)

Try it on your real workload. New accounts start with $5 of credit for Gemma and transcription. DeepSeek, GLM and Qwen3.8 need a top-up.

[Get started](https://inference.broadnet.ai/account/?utm_source=broadrouter&utm_medium=referral&utm_campaign=blog-gdpr-llm-api-checklist&utm_content=page-cta) ·

**Sources**

-   [Regulation (EU) 2016/679 (GDPR), official text on EUR-Lex (fetched 2026-11-14)](https://eur-lex.europa.eu/eli/reg/2016/679/oj)
-   [EDPB Support Pool of Experts, AI Privacy Risks and Mitigations: Large Language Models, 10 April 2025 (fetched 2026-11-14)](https://www.edpb.europa.eu/our-work-tools/our-documents/support-pool-experts-projects/ai-privacy-risks-mitigations-large_en)
-   [EDPB Opinion 28/2024 on personal data in AI models (December 2024; fetched 2026-11-14)](https://edpb.europa.eu/system/files/2024-12/edpb_opinion_202428_ai-models_en.pdf)
-   [European Commission, adequacy decisions including the EU-US Data Privacy Framework (fetched 2026-11-14)](https://commission.europa.eu/law/law-topic/data-protection/international-dimension-data-protection/adequacy-decisions_en)
-   [European Commission, Standard Contractual Clauses (fetched 2026-11-14)](https://commission.europa.eu/law/law-topic/data-protection/international-dimension-data-protection/standard-contractual-clauses-scc_en)
-   [Data Privacy Framework List (US Department of Commerce; fetched 2026-11-14)](https://www.dataprivacyframework.gov/list)
-   [BroadNet Inference privacy statement (effective September 2026; fetched 2026-11-14)](https://www.broadnet.ai/inference-privacy)
