Which AI providers don't train on your API data?
As of 21 November 2026, of 16 AI model offerings tracked, 13 have clear public evidence, 1 is a confirmed no, and 2 have no public evidence.
- Clear public evidence: OpenAI API, Azure OpenAI Service, Anthropic API, Claude via AWS Bedrock, Claude via Google Vertex AI, Gemini via Vertex AI, AWS Bedrock (platform), Mistral via Azure AI, Cohere via AWS Bedrock, Llama via AWS Bedrock, Llama via Azure AI, xAI API, DeepSeek via Fireworks AI
- Confirmed not offered: Cohere API
- No public evidence: Mistral La Plateforme, DeepSeek API (first-party)
Public commitments not to train on customer API data by default, cell by cell, with sources. The cell answers: Is there a public commitment not to train on customer API data by default?
Statuses below are evidence grades, not endorsements, “no public evidence”
means we could not verify it from public sources, not that the answer is no.
●Yes, public
confidence: high · verified 2026-11-16
OpenAI’s own help article publicly states that API data is not used for model training by default, satisfying the commitment.
source · full cell
●Yes, public
confidence: high · verified 2026-11-17
The Azure OpenAI data‑privacy page publicly states that prompts and completions are not used to train foundation models without explicit customer permission, confirming a public commitment.
source · full cell
●Yes, public
confidence: high · verified 2026-11-18
The Anthropic privacy page publicly states a default commitment not to use API data for model training, confirming the claim.
source · full cell
●Yes, public
confidence: high · verified 2026-11-18
The AWS Bedrock FAQ publicly states that customer content is not used to improve base models, confirming a default commitment not to train on API data for Claude via Bedrock.
source · full cell
●Yes, public
confidence: high · verified 2026-11-20
Anthropic’s own privacy page publicly states that for commercial products, including the Anthropic API used by Claude via Vertex AI, inputs and outputs are not used to train models by default.
source · full cell
●Yes, public
confidence: high · verified 2026-11-18
The Google-owned page states verbatim the Training Restriction commitment not to train/fine-tune on customer data without prior permission, covering all managed models (GA and pre-GA) on Gemini Enterprise Agent Platform (formerly Generative AI on Vertex AI), making the public yes_public value correc
source · full cell
●Yes, public
confidence: high · verified 2026-11-21
The provider's public privacy page explicitly states that customer prompts and outputs are not used for training by default.
source · full cell
○No public evidence
confidence: high · verified 2026-11-19
Mistral states data may be used for training by default and offers opt‑out, providing no public commitment not to train on customer API data.
source · full cell
●Yes, public
confidence: high · verified 2026-11-17
The Azure Foundry data‑privacy page publicly states that prompts and outputs are not used to train or improve provider models, applying to Mistral as a model sold by Azure.
source · full cell
Cohere API
Cohere model, served by Cohere (first-party)
✕No, verified
confidence: high · verified 2026-11-16
Cohere's own enterprise data commitments page (cohere.com/enterprise-data-commitments) documents an opt-out toggle for training on prompts/generations ("You can opt out from your prompts and generations being used to train Cohere models in your dashboard settings at any time", "Adjust the toggle to
source · full cell
●Yes, public
confidence: high · verified 2026-11-16
Cohere’s Enterprise Data Commitments page publicly states they do not receive customer prompts or generations on third‑party cloud AI/ML platforms, confirming a default commitment not to train on that API data.
source · full cell
●Yes, public
confidence: high · verified 2026-11-20
The Bedrock FAQ publicly states that customer content is not used to improve base models nor shared, confirming a default no‑training‑on‑API‑data commitment for Llama via Bedrock.
source · full cell
Llama via Azure AI
Meta model, served by Microsoft Azure (Azure AI Foundry / Models-as-a-Service)
●Yes, public
confidence: high · verified 2026-11-20
The Microsoft security blog publicly states that customer data is not used to train models, providing an unrestricted public commitment for Llama via Azure AI.
source · full cell
xAI API
xAI model, served by xAI (first-party)
●Yes, public
confidence: high · verified 2026-11-16
The provider's own security FAQ publicly states they never train on API data without explicit permission, confirming a default commitment not to train.
source · full cell
○No public evidence
confidence: high · verified 2026-11-19
DeepSeek states user input may be used for training by default with an opt‑out option, indicating no public commitment not to train on API data.
source · full cell
●Yes, public
confidence: high · verified 2026-11-20
The Fireworks AI privacy policy publicly commits that API inputs are not used to train models without explicit opt‑in, satisfying the question.
source · full cell