induwara.lk
induwara.lkAI · Data privacy

AI Data Privacy Comparison: Does ChatGPT, Claude or Gemini Train on Your Data?

Before you paste a client's code, a contract or a CV into an AI, check whether it becomes training data. Compare every major provider across the consumer app, the API, and enterprise tiers — training, retention, human review, opt-out, and Zero-Data-Retention — each cell cited to the official policy. No signup, sources below.

By Induwara AshinsanaUpdated Jun 22, 2026
Does it train on your data?19 services
Policies verified 2026-06-22
Verdict19 services shown

Safe with zero configuration: OpenAI API (default), ChatGPT Enterprise / Team, Anthropic API (Commercial), Claude Enterprise, Gemini API (paid) / Vertex AI, Gemini for Google Workspace, Microsoft 365 Copilot (work), xAI API (developers), La Plateforme (API).

9
No training
6
Opt-out
4
Trains by default
ChatGPT (Free & Plus)
OpenAI · Consumer app
Opt-out
Retention
Deleted chats removed within 30 days; kept while account active otherwise
Human review
Limited
Opt-out
Yes
Zero-data-retention
No
Certifications
SOC 2 Type 2

How to opt out: Settings → Data Controls → turn off “Improve the model for everyone.” Temporary Chat also excludes a conversation from training.

OpenAI — How your data is used to improve model performance
OpenAI API (default)
OpenAI · API
No
Retention
Up to 30 days for abuse monitoring, then deleted; 0 days with ZDR
Human review
Limited
Opt-out
Yes
Zero-data-retention
Yes
Certifications
SOC 2 Type 2, GDPR, CCPA
OpenAI — Enterprise privacy & API data usage
ChatGPT Enterprise / Team
OpenAI · Enterprise
No
Retention
Admin-controlled retention; not used for training
Human review
No
Opt-out
Yes
Zero-data-retention
Yes
Certifications
SOC 2 Type 2, GDPR, CCPA
OpenAI — Enterprise privacy
Claude (Free, Pro & Max)
Anthropic · Consumer app
Opt-out
Retention
Up to 5 years if you allow training; ~30 days if you decline
Human review
Limited
Opt-out
Yes
Zero-data-retention
No
Certifications
None listed

How to opt out: Settings → Privacy → turn off “Help improve Claude.” Anthropic prompts new and existing consumer users to choose.

Anthropic — Privacy Center (consumer data & training)
Anthropic API (Commercial)
Anthropic · API
No
Retention
Inputs/outputs deleted within ~30 days unless flagged; 0 days with ZDR
Human review
No
Opt-out
Yes
Zero-data-retention
Yes
Certifications
SOC 2 Type 2, HIPAA-eligible
Anthropic — Commercial Terms of Service
Claude Enterprise
Anthropic · Enterprise
No
Retention
Admin-set retention; not used for training
Human review
No
Opt-out
Yes
Zero-data-retention
Yes
Certifications
SOC 2 Type 2, HIPAA-eligible
Anthropic — Commercial Terms / Enterprise
Gemini app (consumer)
Google · Consumer app
Opt-out
Retention
Human-reviewed samples kept up to 3 years, even after you delete chats
Human review
Yes
Opt-out
Yes
Zero-data-retention
No
Certifications
None listed

How to opt out: Gemini Apps → Activity → turn off “Gemini Apps Activity.” Note: a reviewed sample is still retained up to three years.

Google — Gemini Apps & your data
Gemini API (paid) / Vertex AI
Google · API
No
Retention
Not used to train; retained only for abuse monitoring per terms
Human review
No
Opt-out
Yes
Zero-data-retention
Yes
Certifications
SOC 1/2/3, ISO 27001, HIPAA-eligible (Vertex AI)
Google — Gemini API additional terms
Google AI Studio (free / unpaid)
Google · API
Yes
Retention
Used to improve Google products; human-reviewed samples retained
Human review
Yes
Opt-out
No
Zero-data-retention
No
Certifications
None listed
Google — Gemini API terms (unpaid services)
Gemini for Google Workspace
Google · Enterprise
No
Retention
Stays within your Workspace data governance; not used to train
Human review
No
Opt-out
Yes
Zero-data-retention
No
Certifications
SOC 1/2/3, ISO 27001, ISO 27701, HIPAA-eligible
Google Workspace — Gemini data protection
Copilot (consumer)
Microsoft · Consumer app
Opt-out
Retention
Interactions retained per Microsoft privacy statement
Human review
Limited
Opt-out
Yes
Zero-data-retention
No
Certifications
None listed

How to opt out: Copilot → Settings → Privacy → turn off model-training of your conversations (where the toggle is offered in your region).

Microsoft — Copilot data, privacy & security
Microsoft 365 Copilot (work)
Microsoft · Enterprise
No
Retention
Stays in your tenant; not used to train foundation models
Human review
No
Opt-out
Yes
Zero-data-retention
No
Certifications
SOC, ISO 27001, GDPR, EU Data Boundary
Microsoft — Microsoft 365 Copilot privacy
Meta AI (WhatsApp, Instagram, Facebook)
Meta · Consumer app
Yes
Retention
Retained and used to train generative-AI models per Meta policy
Human review
Yes
Opt-out
No
Zero-data-retention
No
Certifications
None listed

How to opt out: No global training opt-out for chats. In the EU/UK you can submit Meta's “object to your information being used for AI” form; elsewhere the option is limited.

Meta — Generative AI & your information
Grok (consumer, via X)
xAI · Consumer app
Opt-out
Retention
Retained and used for training unless you disable data sharing
Human review
Limited
Opt-out
Yes
Zero-data-retention
No
Certifications
None listed

How to opt out: X → Settings → Privacy and safety → Grok & third-party collaborators → turn off using your posts and interactions to train Grok.

xAI — Privacy policy
xAI API (developers)
xAI · API
No
Retention
Business/API data is not used to train models per xAI terms
Human review
No
Opt-out
Yes
Zero-data-retention
No
Certifications
None listed
xAI — Privacy policy (business data)
DeepSeek (app & web)
DeepSeek · Consumer app
Yes
Retention
Stored on servers in the People's Republic of China; used to improve services
Human review
Yes
Opt-out
No
Zero-data-retention
No
Certifications
None listed
DeepSeek — Privacy policy
DeepSeek API
DeepSeek · API
Yes
Retention
Stored in China; retention and training per the same privacy policy
Human review
Yes
Opt-out
No
Zero-data-retention
No
Certifications
None listed
DeepSeek — Privacy policy
Le Chat (consumer)
Mistral · Consumer app
Opt-out
Retention
Retained and may improve models unless you opt out; EU-hosted
Human review
Limited
Opt-out
Yes
Zero-data-retention
No
Certifications
GDPR

How to opt out: Le Chat → Settings → Data & privacy → turn off using your conversations to improve Mistral's models.

Mistral — Privacy policy
La Plateforme (API)
Mistral · API
No
Retention
Not used to train unless you opt in; EU data hosting
Human review
No
Opt-out
Yes
Zero-data-retention
No
Certifications
GDPR
Mistral — Privacy policy (API)

Every cell is a point-in-time fact from each provider's own policy, last verified 2026-06-22. Tap a source link to confirm the current terms — policies change without notice.

How it works

This is a curated reference, not a calculator. Each row is one provider on one access tier, and every classification is copied from that provider's own published privacy policy or terms of service — never our opinion. The single question it answers is the one most people actually Google: if I type this in, does it train the model?

The key insight the table makes obvious is that the tier matters more than the brand. The same company can train on your data in its free consumer app yet contractually promise never to train on it through its paid API. So each service is classified on three things:

  • Trains on your data? no (safe with zero configuration), opt-out (on by default, but a documented toggle stops it), or yes (trains with no general opt-out).
  • Retention, human review, opt-out and Zero-Data-Retention (ZDR) — the supporting facts that decide whether an opt-out is enough for your risk level. ZDR means inputs are never stored, the strongest guarantee.
  • Certifications — SOC 2, ISO 27001, GDPR and HIPAA-eligible flags, where the provider publishes them for that tier.

Filtering and the verdict are pure, deterministic derivations over that data — no network call, no scoring weights. “Most private only” keeps a row when trainsOnData = no OR (a documented opt-out AND a ZDR path). The verdict card simply partitions the visible rows into no training, opt-out, and trains by default, and names the safe-by-default services.

Every classification is double-checked before the page renders: an internal integrity test confirms that no service is marked both “trains” and “zero-data-retention,” that each “opt-out” row really has an opt-out path, and that two independent derivations of “safe by default” agree. Because the data is a dated snapshot (2026-06-22), each row links its live policy so you can confirm the current terms — policies change without notice.

Worked examples

A freelancer about to paste client code (filter: API & developers)

  1. OpenAI API (default): Trains on data — NO (Up to 30 days for abuse monitoring, then deleted; 0 days with ZDR)
  2. Anthropic API (Commercial): Trains on data — NO (no training on inputs or outputs)
  3. Gemini API (paid) / Vertex AI: Trains on data — NO
  4. Google AI Studio (free / unpaid): Trains on data — YES — avoid for confidential code
  5. Verdict: safe by default = OpenAI API, Anthropic API, Gemini paid API; avoid free AI Studio.

A student worried about the free chatbots (filter: Personal chatbot)

  1. ChatGPT (Free & Plus): OPT-OUT — toggle off “Improve the model for everyone”
  2. Claude (Free, Pro & Max): OPT-OUT — toggle off “Help improve Claude” (allowing it extends retention to 5 years)
  3. Gemini app (consumer): OPT-OUT — turn off Gemini Apps Activity (a sample is still kept up to 3 years)
  4. Verdict: none are safe with zero config, but every one shown can be switched off — follow the opt-out steps first.

Edge case — “Most private only” across every tier

  1. Rule: keep a row if trainsOnData = no OR (opt-out available AND zero-data-retention).
  2. All API/enterprise “No” rows qualify (e.g. OpenAI API, Anthropic API, Gemini paid API).
  3. No consumer chatbot has ZDR, so none of the “opt-out” apps slip through.
  4. Result: the list never contains a free consumer chatbot — exactly the safe set you'd hand a security reviewer.

Frequently asked questions

Sources & references

Every row links its own policy in the tool above. The primary sources, by provider:

All 19 classifications were last cross-checked against these official policies on 2026-06-22. The data is a dated snapshot, refreshed as policies change. This page is informational and is not legal advice or a substitute for a Data Processing Agreement; for self-hosted open-weight models (Ollama, LM Studio) nothing leaves your machine, a different trust model not covered here. Provider and product names are trademarks of their respective owners.

Related tools

Rate this tool
Be the first to rate

Comments & feedback

Spotted a bug or want an improvement? Tell us — our team reviews every comment, and good ideas get built. Comments are public and anonymous.

Spotted a policy that changed, or a service that should be added?

Email me at [email protected] — most fixes ship within 24 hours.