RETROSPECTIVE RECORD · PREPARED 16 SEPTEMBER 2026Start here · 100 retrospective records ↗
AI Use Field Guide

Start here / Privacy & safety

Privacy & safety / Start-here guide · Start-here guide · prepared 16 September 2026

A business AI plan is a contract, not just a bigger toggle

OpenAI, Anthropic and Google document contractual no-training defaults for paid business tiers that differ from consumer settings.

Visual for this record: A business AI plan is a contract, not just a bigger toggle
Visual published by lh3.googleusercontent.com, shown for identification of the record. Credit: lh3.googleusercontent.com · source page ↗ Rights: owner-review-pending.

Start here

A small team choosing between a personal ChatGPT, Claude or Gemini account and a paid business version is choosing between a setting and a contract. That distinction matters more than price, because a setting can be a default a company chooses to offer, while a contract term is something the company can be held to.

What the documents say

OpenAI's enterprise privacy page, as retrieved on 16 September 2026, states 'we do not train our models on your data by default' for its business products, and that a customer controls how long data is retained on ChatGPT Enterprise, ChatGPT for Healthcare and ChatGPT Edu, a control the same page frames as unavailable on free or Plus consumer tiers. Anthropic's Commercial Terms of Service put the same commitment into contract language rather than marketing copy: 'Anthropic may not train models on Customer Content from Services,' alongside a clause that the customer retains rights to its inputs and owns its outputs. Google's Workspace AI privacy page answers the question directly for business customers: 'Google does not use customers' Workspace data to train or improve the underlying generative AI and large language models... that power Gemini, Search, and other systems... without permission.' All three describe a default opposite to at least part of each vendor's own consumer product, where training can be the default unless a person opts out.

Check this

The check worth running before paying for a business tier is where the promise actually lives: a public web page describing a default, or a signed agreement's specific clause. Anthropic's commitment appears in the terms document itself, the strongest form, since it is contractually binding. OpenAI's and Google's statements appear on marketing and security pages rather than quoted contract text, so a buyer with real sensitivity should ask which specific agreement, such as a data processing addendum, carries the no-training promise, and read that document rather than the public page alone.

What holds and what fails

What holds across all three is the consistent shape of the offer: a business tier converts a togglable default into a stated commitment, and typically adds retention controls a personal account lacks. What this does not establish is that every feature or workspace setting automatically inherits that promise; each page describes its scope by specific product names, and a feature outside that scope, or a personal account signed in at a workplace, would not be covered. This is an editorial reading of promotional and contractual pages together, a starting point for a reader's own verification, not legal advice.

  • Ask which specific document, not just which web page, carries your provider's no-training promise.
  • Check that the business tier's promise names the product you are actually using.
  • Do not assume a personal account inherits a workplace's business-tier protections.

A business plan is worth its price partly because a promise printed in a contract is a different kind of thing than a promise printed on a marketing page, even when the words sound nearly identical.

Sources & reading trail

Enterprise privacy at OpenAI ↗

States that OpenAI does not train on enterprise customer data by default and that customers control retention on named enterprise products.

Source published: Not established · Retrieved: 16 September 2026

Commercial Terms of Service ↗

Contains the contractual clause that Anthropic may not train models on Customer Content from its commercial Services.

Source published: Not established · Retrieved: 16 September 2026

Generative AI Security, Compliance and Privacy ↗

States that Google does not use Workspace customer data to train or improve the underlying models that power Gemini.

Source published: Not established · Retrieved: 16 September 2026

Documentation, regulator guidance and studies establish the record; the checks and the boundary are AI Use Field Guide editorial analysis. This retrospective draft does not imply the site published on the event date.