OpenAI Bars Rival Training, Vets Model Resellers From Nov 20

OpenAI’s updated Usage Policies take effect on 20 November 2026. They require any service that lets third parties directly access OpenAI models to get OpenAI’s prior review and approval, and they bar using OpenAI models or outputs to build training material for models meant to replace OpenAI’s.

OpenAI’s changelog dates the update to 21 October 2026, which by our arithmetic gave 30 days’ notice; the rules apply in 5 days. The policy text grew from 943 words in the previous version to 1,553 words, about 65% longer by our count.

Business Pill · Switching Costs Explained

An evergreen Business Engineer explainer on switching costs. It teaches the general idea only and does not discuss this policy.

What Changes for Resellers and Gateways

Under a new heading, “Providing access to our models”, the policy says: “If you operate a model-access intermediary platform or other service that enables third parties to directly access OpenAI models, you must obtain our prior review and approval.” The previous version of the policy, dated 29 October 2025, did not use the word intermediary. OpenAI’s help center says that until 20 November “the previous version remains in effect”.

OpenAI draws a line for ordinary apps: “This requirement does not apply solely because an application integrates OpenAI models to power its features or functionality.” Separately, OpenAI’s Services Agreement already says a customer “may not resell or lease access to its Account or any End User Account”.

A post on X on 15 November described the change as cutting off unapproved model routers, API relays and AI gateways. The policy text itself sets an approval requirement; it does not describe how or when unapproved services would lose access.

The Clarified Limit on Training Rival Models

The policy says: “You must not use OpenAI models or their outputs to generate, prepare, select, or supply training material or signals for models that are intended as alternatives to or substitutes for OpenAI models or services.” OpenAI’s help-center note on the update describes this as having “Clarified our existing restrictions on using OpenAI models and outputs to train models intended as alternatives to or substitutes for OpenAI models or services”. The policy lists three covered activities: preparing training data, tasks, rubrics or reinforcement-learning environments; providing feedback such as labels, scores, rankings and rewards; and selecting, filtering or validating training data.

The policy adds that the restrictions apply “whether you perform the work yourself, for another party, or through someone acting on your behalf, even if you do not train the resulting model”. Users also “must not attempt to extract protected reasoning through unauthorized access, reverse engineering, or circumvention of safeguards”. Research, benchmarking and product development remain allowed “within these boundaries”, the policy says. In October we reported that OpenAI said it disrupted a distillation campaign.

What Else the Update Covers

The changelog says the update also “clarified protections for children and teens and requirements for high-stakes uses”, and that it “incorporates our existing national security principles and sets out approval requirements for national security and law enforcement uses, as well as for model-access intermediaries”.

The Structural Read

As we read the training clause, it covers the supply chain around a rival model, not only the training run: the policy lists preparing data, giving feedback and filtering data, and applies even if you do not train the resulting model.

As we read the timing, OpenAI gave 30 days between the changelog entry on 21 October and the effective date on 20 November, and the policy text does not describe how approval or enforcement will work.

Three Implications

FOR MODEL ROUTERS AND API GATEWAYS The policy says a service that enables third parties to directly access OpenAI models must obtain OpenAI’s prior review and approval.

FOR APP BUILDERS The policy says the approval requirement does not apply solely because an application integrates OpenAI models to power its features.

FOR LABS AND DATA VENDORS The policy bars using OpenAI models or outputs to prepare training material for models intended as alternatives to OpenAI’s.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Open vs Closed Meta-Framework.

The framework states the pattern as: “Close the SCARCE layer (keep proprietary) + Open the ABUNDANT layer (commoditize)”.

As we read it, the policy closes access to the model layer through intermediaries while leaving app-building open; the framework does not discuss this policy.

What Is Not Established

OpenAI’s policy page does not say how many intermediary services exist, how the review works, how long it takes, or what happens to services that are not approved by 20 November. We found no separate OpenAI announcement on enforcement, and OpenAI was not contacted for this piece.

The Bottom Line

From 20 November 2026, OpenAI’s Usage Policies require prior review and approval for any service that gives third parties direct access to its models. They also bar using its outputs to train models meant as alternatives to its own, according to the policy text dated 21 October.