Model

OpenAI requires approval for platforms reselling its models, effective November 20

OpenAI's revised usage policies require prior approval for any platform reselling direct access to its models, effective November 20. The middleware the market just valued in the billions gets a month to decide, and its own plumbing is already cracking.

Richard TangRichard Tangtwo-minute read
Share
Sam Altman, OpenAI's chief executive, with Masayoshi Son in Tokyo
1 / 5Slide 1 of 5
OpenAI chief executive Sam Altman in Tokyo in February 2025. His company's revised usage policies put an approval checkpoint in front of the model-routing layer.

OpenAI$1.18T — OpenAI, private, latest valuation $1.18T has rewritten its usage policies so any platform offering third parties direct access to its models must first win OpenAI's review and approval, effective November 20, 2026 1. The revision is dated October 21, so the routing layer has four weeks to pick a side 12.

The rule draws a clean line: an app that uses OpenAI models to power its own features is untouched, while a platform that redistributes access is now, in effect, a licensee 12. The biggest lab just recaptured the direct customer relationship and put an enforcement checkpoint in front of the middleman business.

Why the middleman became the problem

When a developer connects directly, OpenAI knows exactly who it is serving. Through a gateway, the end user sits one step removed, and rules on protecting minors, restricting high-stakes decisions and governing national-security use were getting hard to enforce; the approval process gives OpenAI a look at the platform before it resells 2.

The gateways are also where OpenAI's rivals now distribute: TypeSafe's decision model Jev became the fastest-adopted model ever on Vercel's AI Gateway, reaching about 13% of paid teams within a day of its September launch 3. Gating the rails is gating the competition's distribution.

The layer just learned its margins are conditional

The operators pay first: OpenRouter, which Stripe has reportedly agreed to buy for more than $7 billion, and the LiteLLM-based stacks enterprises self-host 45. Miss the November date and an unapproved gateway's customers can lose their access path to OpenAI 2. The layer's plumbing is fragile in its own right: Respan scrapped more than 100,000 lines of wrapper code around LiteLLM after a January upgrade broke connections it never root-caused ("We still haven't figured out the cause") 6. Young software, dependent on the labs above and its own dependencies below.

Approval is not the end of the gateway

The value that survives is value beyond resale. A gateway that routes on cost and capability, hosts MCP tools behind its own credentials, and meters every request does work the model APIs do not 7. Those gateways will win approval and keep their margin; pure relayers will get cut. Reselling choice is no longer a feature. It is a license, and the teams who learn this routing layer now, as the plumbing matures, are building on the right side of the gate.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Reader comments

0 comments

    Model

    Anthropic hired a UK lobby firm the week it priced its $1.4 trillion IPO

    The same week Anthropic set its $70-to-78 IPO range, the UK lobbyist registry showed it paying for political access in three countries. Goldman's $685 billion revenue case runs on data centers that voters and regulators must permit, and buyers fund that insurance without seeing its cost.

    Model

    Mistral built its enterprise stack in eight days and priced none of it

    Between October 29 and November 5 the Paris lab published its biggest model's weights, expanded its agent platform and bought a fifth startup, all funded by a September mark of €21 billion. Not one piece of the stack carries a published price, while free Chinese open-weight models still outscore it.

    Model

    xAI's Grok 4.8 entered training 57 days late with no price card; Anthropic pays SpaceX $1.25 billion a month

    The 57-day wait bought a serving stack xAI wrote itself, and the delay ends with no scores and no price. For a SpaceX shareholder, the unpublished price card for 4.8 is the next receipt.

    Model

    Meta's Muse Ember costs $0.14 a task, a third of Claude Haiku, as Anthropic prices its IPO

    Meta's smallest model does the work agents repeat all day at $0.14 a task, matching Anthropic's Haiku at a third of the cost. Anthropic prices its IPO this week on about $81 billion of run-rate token revenue, and the floor under that business just moved.

    Model

    Anthropic cuts Fable 5.2 prices 20% on IPO roadshow eve, and Google and Broadcom shareholders carry the bill

    Anthropic discounted its flagship the day its IPO quiet period ended, for buyers who had already stopped climbing. The compute bill behind the discount is owed chiefly to Google and chipmaker Broadcom, which leaves their shareholders carrying the exposure.

    Model

    AfterQuery's agent training data lifted OpenAI's small model from 3.1% to 17% on Terminal-Bench

    The November 4 run, showcased on Thinking Machines' Tinker, sits beside rival Mercor's proof on the same page, giving labs a direct comparison of measured benchmark uplift for the first time. It is the claim beneath AfterQuery's reported $3.2 billion valuation, now competing in public for labs' data budgets.

    Get the newsletter

    The day's AI and market news, in your inbox every weekday.