ChatGPT's business tiers default away from training use

The writer's problem

A writer choosing among ChatGPT's plans for drafting or research is choosing more than a monthly bill: the plans also differ in what happens to the conversations themselves. OpenAI's own pricing page and enterprise privacy page, as retrieved on 16 September 2026, describe both the usage tiers and the data-handling defaults attached to each.

What the documents show

The pricing page lists four plans aimed at individuals, Free, Go, Plus, and Pro, alongside Business and Enterprise plans for organizations, with Business available starting at 2 users. A feature comparison on the same page states, identically for Free, Go, Plus, and Pro, that content is used to train OpenAI's models, with opt-out available through account settings; none of the four consumer tiers include administrative features such as SAML single sign-on or a listed SOC 2 Type 2 compliance status, which the page reserves for the organizational plans. The enterprise privacy page, last updated 8 January 2026, states a different default for Business, Enterprise, Healthcare, Edu, and Teachers accounts: by default, we do not use your business data for training our models, adding that shared data is used for training only if you have explicitly opted in. The same page states that Enterprise, Healthcare, and Edu customers additionally control how long their data is retained, a control the pricing page does not extend to the individual plans.

The editorial choice

This is an editorial judgment rather than a claim either page makes for the reader: a writer drafting confidential material, an unpublished manuscript under contract, a client's unreleased material, or a source's private account, should weigh whether an opt-out on a personal plan is an acceptable safeguard or whether the default-off training policy attached to a business-tier account better matches the sensitivity of the material, independent of which plan is cheaper.

What stays with the author

Neither page states what happens to material after a writer copies a chatbot's output into a manuscript, and neither addresses whether the facts or phrasing in a response are accurate. Setting the opt-out, if a personal plan is used, remains an action the account holder must take; the default the page states applies only until changed.

• Has the training opt-out been checked and confirmed inside account settings, or is the default still in effect?
• Does the material being drafted warrant the business-tier default of no training rather than a consumer plan's opt-out model?
• Do the plan's usage limits, not just its price, match how the manuscript work is actually being done?

A pricing page that states a default in plain terms is more useful to a writer than one that only advertises features, because the default is the setting most people never change.