vendor_id stringlengths 10 22 | vendor_name stringclasses 8
values | product stringlengths 10 30 | tier stringclasses 4
values | field stringclasses 4
values | value stringclasses 6
values | quote stringlengths 50 193 | source_url stringlengths 22 116 | confidence stringclasses 3
values | checked_at stringdate 2026-08-13 00:00:00 2026-08-14 00:00:00 |
|---|---|---|---|---|---|---|---|---|---|
aws-bedrock | Amazon | Amazon Bedrock | enterprise | opt_out_available | not_applicable | Not used for foundation-model training by default; model providers cannot access prompts and completions. | https://docs.aws.amazon.com/bedrock/latest/userguide/data-protection.html | medium | 2026-08-14 |
aws-bedrock | Amazon | Amazon Bedrock | enterprise | trains_on_your_data | no | Because the model providers don't have access to those accounts, they don't have access to Amazon Bedrock logs or to customer prompts and completions. | https://docs.aws.amazon.com/bedrock/latest/userguide/data-protection.html | medium | 2026-08-14 |
aws-bedrock | Amazon | Amazon Bedrock | enterprise | zero_retention_available | yes | Zero data retention. No request or response data is written to durable storage by AWS or shared with the model provider. | https://docs.aws.amazon.com/bedrock/latest/userguide/data-retention.html | high | 2026-08-14 |
anthropic-api | Anthropic | Claude API | api | opt_out_available | not_applicable | Customer Content is not used for training by default, so no opt-out is required. | https://www.anthropic.com/legal/commercial-terms | high | 2026-08-13 |
anthropic-api | Anthropic | Claude API | api | trains_on_your_data | no | Anthropic may not train models on Customer Content from Services. | https://www.anthropic.com/legal/commercial-terms | high | 2026-08-13 |
anthropic-api | Anthropic | Claude API | api | zero_retention_available | unclear | The Commercial Terms confirm no training on Customer Content but contain no explicit zero-data-retention statement; marked unclear pending a direct source. | https://www.anthropic.com/legal/commercial-terms | medium | 2026-08-13 |
github-copilot | GitHub | GitHub Copilot Business | enterprise | opt_out_available | not_applicable | Business/Enterprise data is not used for training by default. | https://github.blog/changelog/2026-03-25-updates-to-our-privacy-statement-and-terms-of-service-how-we-use-your-data/ | high | 2026-08-13 |
github-copilot | GitHub | GitHub Copilot Business | enterprise | trains_on_your_data | no | GitHub does not use either Copilot Business or Enterprise data to train its models (the 2026-03-25 training change applies only to Free/Pro/Pro+). | https://github.blog/changelog/2026-03-25-updates-to-our-privacy-statement-and-terms-of-service-how-we-use-your-data/ | high | 2026-08-13 |
github-copilot | GitHub | GitHub Copilot Business | enterprise | zero_retention_available | yes | For Copilot Business and Enterprise, prompts and suggestions are not retained. (Re-pull exact clause via verifier.) | https://github.blog/changelog/2026-03-25-updates-to-our-privacy-statement-and-terms-of-service-how-we-use-your-data/ | medium | 2026-08-13 |
google-gemini-consumer | Google | Gemini app (consumer) | consumer | opt_out_available | yes | You can change your auto-delete setting in Gemini Apps Activity from the default of 18 months to 3 months, 36 months, or indefinite. You can also manually delete your Gemini Apps chats anytime. | https://support.google.com/gemini/answer/13594961?hl=en | medium | 2026-08-14 |
google-gemini-consumer | Google | Gemini app (consumer) | consumer | trains_on_your_data | opt_out_default_on | Please don't enter confidential information that you wouldn't want a reviewer to see or Google to use to improve our services, including machine-learning technologies. | https://support.google.com/gemini/answer/13594961?hl=en | medium | 2026-08-14 |
google-gemini-consumer | Google | Gemini app (consumer) | consumer | zero_retention_available | unclear | No consumer zero-retention option; chats reviewed by human reviewers are retained for up to three years. | https://support.google.com/gemini/answer/13594961?hl=en | low | 2026-08-14 |
google-vertex | Google | Vertex AI (Gemini, enterprise) | enterprise | opt_out_available | not_applicable | Customer data is not used to train foundation models by default on the enterprise tier. | https://cloud.google.com/vertex-ai/generative-ai/docs/data-governance | high | 2026-08-13 |
google-vertex | Google | Vertex AI (Gemini, enterprise) | enterprise | trains_on_your_data | no | Google won't use your data to train or fine-tune any AI/ML models without your prior permission or instruction. | https://cloud.google.com/vertex-ai/generative-ai/docs/data-governance | high | 2026-08-13 |
google-vertex | Google | Vertex AI (Gemini, enterprise) | enterprise | zero_retention_available | yes | Google Cloud publishes a dedicated "Generative AI and zero data retention" policy for Vertex AI; ZDR is offered for eligible use. (Re-pull exact clause via verifier.) | https://cloud.google.com/vertex-ai/generative-ai/docs/data-governance | medium | 2026-08-13 |
meta-llama | Meta | Llama (open-weight models) | open-weight | opt_out_available | unclear | Governed by the host offering, not by Meta; see the specific host (Bedrock, Azure, etc.). | https://www.llama.com/ | low | 2026-08-14 |
meta-llama | Meta | Llama (open-weight models) | open-weight | trains_on_your_data | unclear | Llama is an open-weight model; data handling depends on the host you run it on (e.g., AWS Bedrock, Azure, or self-hosting), not on Meta. | https://www.llama.com/ | low | 2026-08-14 |
meta-llama | Meta | Llama (open-weight models) | open-weight | zero_retention_available | unclear | Depends on the host that runs the model; Meta does not receive inputs or outputs from third-party or self-hosted deployments. | https://www.llama.com/ | low | 2026-08-14 |
azure-openai | Microsoft | Azure OpenAI Service | enterprise | opt_out_available | not_applicable | are NOT used to train any generative AI foundation models without your permission or instruction. | https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/openai/data-privacy | high | 2026-08-14 |
azure-openai | Microsoft | Azure OpenAI Service | enterprise | trains_on_your_data | no | prompts and completions are not used to train, retrain, or improve the base models. | https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/openai/data-privacy | high | 2026-08-14 |
azure-openai | Microsoft | Azure OpenAI Service | enterprise | zero_retention_available | yes | the data storage and human review process described above is not performed. | https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/openai/data-privacy | medium | 2026-08-14 |
microsoft-copilot | Microsoft | Microsoft 365 Copilot | enterprise | opt_out_available | not_applicable | Not used for foundation-model training by default. | https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-privacy | high | 2026-08-13 |
microsoft-copilot | Microsoft | Microsoft 365 Copilot | enterprise | trains_on_your_data | no | Prompts, responses, and data accessed through Microsoft Graph aren't used to train foundation LLMs, including those used by Microsoft 365 Copilot. | https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-privacy | high | 2026-08-13 |
microsoft-copilot | Microsoft | Microsoft 365 Copilot | enterprise | zero_retention_available | unclear | Not zero-retention: interaction data is stored and admin-controlled — "admins can also use Microsoft Purview to set retention policies for the data related to chat interactions with Copilot." | https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-privacy | medium | 2026-08-13 |
mistral-api | Mistral AI | Mistral API (La Plateforme) | api | opt_out_available | not_applicable | API data is not used for model training by default (note: enabling Labs models allows training regardless of plan). | https://docs.mistral.ai/admin/monitor-comply/privacy-data-controls | high | 2026-08-14 |
mistral-api | Mistral AI | Mistral API (La Plateforme) | api | trains_on_your_data | no | API: data sent through the API isn't used for model training. | https://docs.mistral.ai/admin/monitor-comply/privacy-data-controls | high | 2026-08-14 |
mistral-api | Mistral AI | Mistral API (La Plateforme) | api | zero_retention_available | yes | When zero data retention is enabled, troubleshooting and analytics data are not stored. | https://docs.mistral.ai/admin/monitor-comply/privacy-data-controls | medium | 2026-08-14 |
openai-api | OpenAI | OpenAI API | api | opt_out_available | not_applicable | By default, we do not use your business data for training our models (opt-in required to share). | https://openai.com/enterprise-privacy | high | 2026-08-13 |
openai-api | OpenAI | OpenAI API | api | retention_default | 30 days | OpenAI may securely retain API inputs and outputs for up to 30 days to provide the services and to identify abuse. After 30 days, API inputs and outputs are removed from our systems | https://openai.com/enterprise-privacy | high | 2026-08-13 |
openai-api | OpenAI | OpenAI API | api | trains_on_your_data | no | the API Platform (after March 1, 2023) isn't used for training our models, unless you have explicitly opted in to share your data with us to improve the services | https://openai.com/enterprise-privacy | high | 2026-08-13 |
openai-api | OpenAI | OpenAI API | api | zero_retention_available | yes | You can also request zero data retention (ZDR) for eligible endpoints if you have a qualifying use-case. | https://openai.com/enterprise-privacy | high | 2026-08-13 |
openai-consumer | OpenAI | ChatGPT (consumer) | consumer | opt_out_available | yes | Go to Settings > Data Controls and turn off "Improve the model for everyone." | https://help.openai.com/en/articles/7730893-data-controls-faq | high | 2026-08-13 |
openai-consumer | OpenAI | ChatGPT (consumer) | consumer | trains_on_your_data | opt_out_default_on | Data Controls let you decide how ChatGPT uses your conversations... choose whether your conversations help improve our models (on by default; turn off "Improve the model for everyone"). | https://help.openai.com/en/articles/7730893-data-controls-faq | high | 2026-08-13 |
openai-consumer | OpenAI | ChatGPT (consumer) | consumer | zero_retention_available | unclear | No consumer zero-data-retention option is documented; retention is governed by Data Controls and Temporary Chats (deleted after 30 days). Marked unclear pending a direct statement. | https://help.openai.com/en/articles/7730893-data-controls-faq | low | 2026-08-13 |
AI Vendor Data-Trust Index
A source-cited record of how AI products handle your data — whether the vendor trains on your inputs,
whether zero-data-retention is available, and whether you can opt out — across 11 products. Every value
carries a verbatim quote from the vendor's live policy, the source URL, a confidence level, and the
date it was checked. Ambiguous cases are marked unclear rather than guessed.
⚠️ This is a dated snapshot — the live index is authoritative
Vendor policies change. This file is a point-in-time export. For the current, continuously re-verified data, always use the live source:
- Human view: https://data.oscar-lopez.com/ai-trust
- API (free):
GET https://trust.oscar-lopez.com/v1/ai-trust/matrix - MCP:
https://trust.oscar-lopez.com/mcp
The live index is re-checked weekly by an automated verifier that only publishes a value when a verbatim quote from the policy supports it. This snapshot should be treated as illustrative; verify against the live source and the linked policy before relying on any value.
Snapshot date: 2026-08-14
Columns
| column | meaning |
|---|---|
vendor_id |
stable id, e.g. openai-api, anthropic-api, azure-openai |
vendor_name, product, tier |
vendor, product, and tier (api/enterprise/consumer/open-weight) |
field |
trains_on_your_data, zero_retention_available, opt_out_available, retention_default |
value |
normalized answer: no, yes, opt_out_default_on, not_applicable, unclear, or a string |
quote |
verbatim supporting text from the source (or an explanatory note where value is unclear) |
source_url |
the policy page the quote is from |
confidence |
high / medium / low |
checked_at |
ISO date the value was last verified |
Vendors covered
OpenAI (API + ChatGPT), Anthropic (Claude API), Google (Vertex AI + Gemini app), Microsoft (365 Copilot + Azure OpenAI), Amazon Bedrock, Mistral (API), GitHub Copilot (Business), Meta Llama (open-weight).
Methodology & scope
Values are extracted from vendors' published policies with a verbatim quote per field. "Unclear" is used deliberately — a missing or ambiguous policy is never rounded up to a favorable answer. This is a factual, source-cited summary of public policies, not legal advice; confirm against the linked source and your own contract.
Citation
López, O. E. (2026). AI Vendor Data-Trust Index. https://data.oscar-lopez.com/ai-trust
Maintained by Oscar E. López, Ph.D. (Information Systems, UMass Amherst; research includes online privacy and trust). Independent project; not affiliated with or endorsed by any listed vendor.
- Downloads last month
- 13