How to check if an AI feature uses your data for model training

TL;DR — To tell whether an AI feature is using your data for training, check three places in order: the product setting for data sharing/training, the provider's privacy or AI policy, and your contract or data processing terms. The most common outcome is: training is off for paid business/workspace accounts but still on unless you disable an opt-in toggle or avoid feedback features. Reading time: ~5 min

Goal

When you finish, you will have a written yes/no answer for a specific AI feature: whether your prompts, files, chat history, or feedback can be used to train models, plus a saved screenshot or policy link that proves it.

Prerequisites

• Your login for the product that includes the AI feature
• Access to the workspace, team, or admin settings if the feature is managed at organization level
• The provider's public website or trust/privacy documentation pages
• Your contract, order form, or Data Processing Addendum (DPA; the legal document about data handling), if you have one
• A note with the exact AI feature name you are checking, for example: "AI chat", "meeting summary", or "document assistant"
• Optional: permission to contact your account manager or vendor support if the policy is unclear

Steps

Step 1: Identify the exact AI feature and account type

Open the product and write down the feature name and your account type from the billing or workspace page.

Use this menu path in the provider's dashboard:

Dashboard or Workspace Settings → Billing or Plan
Dashboard or Workspace Settings → AI, Labs, Features, or Privacy

What you should see when this succeeds: the feature has a specific name and your account is clearly labeled as personal, free, pro, business, enterprise, team, or similar.

Step 2: Check the in-product training or data-sharing setting

Look for a setting that explicitly mentions training, model improvement, data sharing, feedback, or provider use of prompts/files.

Use this menu path in the provider's dashboard:

Settings → Privacy
Settings → Data Controls
Settings → AI Features
Settings → Workspace or Organization → Security & Privacy

Look for labels similar to these exact phrases:

Use customer data to improve models
Allow prompts and responses for training
Share data with provider for model improvement
Use content to improve AI services
Include feedback content in training

Record the literal value you see:

On
Off
Enabled
Disabled
Included
Excluded

What you should see when this succeeds: a visible toggle, checkbox, or note that says whether data is used for training or model improvement.

Step 3: Check whether feedback buttons are a separate training path

Many products keep the main feature excluded from training but use thumbs-up, thumbs-down, "Send feedback", or bug-report attachments for model improvement.

Use this menu path:

AI feature screen → Feedback icon
Settings → Privacy → Feedback
Help → Report a problem

Read the text next to the feedback control. If there is a checkbox, record the exact wording.

Common examples:

Include conversation content
Allow attached files to be reviewed
Help improve the model
Share this interaction with the provider

What you should see when this succeeds: either a note that feedback content may be reviewed/used separately, or confirmation that feedback is disabled or excluded.

Step 4: Read the provider's public policy for your account type

Open the provider's website and search for the policy that applies to AI services, privacy, and business data. Use the site's search box first.

Copy-paste these searches into the provider's site search or a web search:

site:provider-domain.example AI training policy
site:provider-domain.example model improvement business data
site:provider-domain.example privacy generative AI enterprise
site:provider-domain.example data processing addendum AI services

You are looking for explicit statements like these:

We do not use customer content to train our models
We may use consumer content to improve services
Business, Team, or Enterprise content is excluded from training by default
Data submitted through feedback may be used for model improvement

What you should see when this succeeds: a policy page that clearly distinguishes consumer/free accounts from business/enterprise accounts, or says training is opt-in/opt-out.

Step 5: Check your contract or DPA for overrides

If you are on a paid workspace, the contract may control data use even if the marketing page is vague.

Open these documents if you have them:

Order Form
Master Services Agreement
Data Processing Addendum
Security Addendum
AI Terms or Product-Specific Terms

Search within the document for these exact terms:

train
training
model improvement
service improvement
customer content
prompts
outputs
feedback
machine learning

What you should see when this succeeds: a clause that either prohibits training on customer content, allows it only with consent, or excludes certain data types while allowing logs/telemetry (usage metadata, such as timestamps and error rates).

Step 6: Map what data types are covered

Do not stop at a single yes/no. Write down which data types are covered, because providers often treat them differently.

Use this checklist and fill in one value for each row:

What you should see when this succeeds: a complete table with no blank rows.

Step 7: Save proof for your team

Take a screenshot of the setting and save the policy URL or contract clause in the same note.

Use this format exactly in your notes app or ticket:

AI feature: <feature name>
Account type: <plan/workspace type>
Training on prompts/files: Yes / No / Unclear
Feedback used separately: Yes / No / Unclear
Policy URL: <full URL>
Contract clause: <document name + section>
Checked on: <YYYY-MM-DD>
Reviewer: <your name>

What you should see when this succeeds: one note or ticket comment that another person can review without repeating your work.

Step 8: If anything is unclear, ask one precise vendor question

If the UI and policy do not match, send a narrow question instead of a broad one.

Copy-paste this message into your vendor support portal or email:

We need a written answer for this specific feature: <feature name>.
For our account type: <plan/workspace type>.
Please confirm whether the following are used for model training or model improvement: prompts, uploaded files, outputs, chat history, feedback submissions, and usage logs.
Please include the policy URL or contract section that applies to our account.

What you should see when this succeeds: a reply that lists data types individually, not a generic privacy statement.

Verify it works

Use these checks to confirm you have a reliable answer:

1. In the product UI, the training/data-sharing toggle is visibly set to Off, Disabled, or equivalent, or the UI says the feature is excluded by default.
2. The public policy page explicitly matches your account type and says whether business/workspace data is used for training.
3. Your note contains all of these fields:

AI feature
Account type
Training on prompts/files
Feedback used separately
Policy URL
Checked on

1. If you have a contract, its wording does not conflict with the UI or policy page.

A simple decision flow looks like this:

[Feature in Product]
      |
      | 1. check toggle
      v
[UI Setting] ---- 2. compare ----> [Public Policy]
      |                                  |
      | 3. save proof                    | 4. compare
      v                                  v
   [Your Note] <---- 5. confirm ---- [Contract/DPA]

Common pitfalls

Mistake: Checking only the marketing page

Symptom: you find a reassuring sentence like "we value privacy" but no direct statement about training. Fix: open the privacy, AI terms, or trust documentation and search for training, model improvement, and customer content.

Mistake: Missing the difference between free/personal and business/workspace accounts

Symptom: the policy seems contradictory because one page says data may be used and another says it is not. Fix: match the wording to your exact account type from Settings → Billing or Plan before you decide.

Mistake: Ignoring feedback and thumbs-up/down controls

Symptom: the main setting says training is off, but shared conversations still leave your workspace through feedback. Fix: open the feedback dialog and record whether Include conversation content or similar is enabled.

Mistake: Assuming "not used for training" means "not stored at all"

Symptom: you tell your team there is zero retention, but the provider still keeps logs for abuse detection or reliability. Fix: separately record usage logs/telemetry as Included, Excluded, or Unclear.

Mistake: Trusting a screenshot without a date or URL

Symptom: later you cannot prove which policy version you checked. Fix: save the full policy URL and add Checked on: YYYY-MM-DD to your note.

Mistake: Asking support a broad question

Symptom: support replies with a generic privacy link that does not answer your case. Fix: send the Step 8 template and ask about prompts, files, outputs, feedback, and logs one by one.

This article was written by an AI system and published pending human review. Verify anything you intend to act on.