OpenAI’s Ads Push Has One Hard Limit: User Privacy

By Sophie Lindqvist

OpenAI posted a $38.5 billion net loss last year, and its advertising test operates under one firm limit: outside partners cannot receive user data.

That pairing gives the nascent operation both a commercial reason and a strict operating boundary. The AI lab wants adtech and agency partners to help scale an ads business that began testing in February, but it will not trade access to ChatGPT conversations for that growth.

Benji Shomair, OpenAI's head of monetization, said user trust ranks as the top priority in building the ads business. In Monday remarks at an event hosted by Luma Partners, an adtech-focused investment bank, he described trust as a guiding principle for product decisions.

The Monday remarks provide no advertising formats, prices, campaign controls, revenue figures, partner names, or launch schedule. They therefore show an early test, not a fully specified commercial offer.

What does this mean for adtech and agency partners?

Partners have room to work with OpenAI, but not through access to the underlying conversations. The company will not share user data, including ChatGPT conversations, with outside ad partners, and it has committed never to sell user data to advertisers. It also says it will keep those conversations private from advertisers.

OpenAI's promise covers both sharing and selling. Those are separate routes for ad companies to obtain user information, and the company says it rejects both. The assurance extends beyond a sales pledge to direct access for outside partners.

That boundary rules out conversation-level data as a routine targeting or measurement input. Shomair did not identify what non-user information OpenAI might supply, how campaigns will run, or whether advertisers can measure results without access to individual exchanges. Partners therefore cannot infer a targeting system from the February test alone.

Shomair said: “There are places where our partners would like us to do certain things that we just can’t because of how we think about making decisions.”

His wording makes the limit a decision rule, not a temporary technical gap. The line separates trust from a negotiable campaign feature: he did not describe privacy as an option available to selected advertisers. Instead, he framed it as a limit on what the company believes it can do at all.

Scale without a data trade

The ads push matters amid OpenAI's cash strain and $38.5 billion net loss last year. The scale gives the company a strong reason to seek another revenue source. The remarks provide no advertising revenue, market forecasts, spending plans, or break-even target, so they do not show how the ads business will affect the loss.

Even so, the financial pressure and the February start point in the same direction. OpenAI is eager to work closely with adtech and agency partners to scale the operation, while the privacy commitment determines how far that cooperation can go.

Neither the size of the loss nor the February test reveals campaign economics. No advertising sales, spending, or return figure appears in Shomair's remarks, leaving agencies and advertisers without enough information to assess the commercial offer.

For agencies, the reported terms leave the operating model unresolved. The remarks do not say which partners joined the test, what each partner will handle, or which ads buyers can select. They also give no timetable for wider rollout and no indication that selling conversation data is an option.

OpenAI is therefore opening the advertising industry without giving advertisers a claim on user data. The company wants commercial scale, but only inside the boundary it has drawn around privacy and trust.

For advertisers, the practical test is whether that model can deliver useful reach and measurement without exposing the conversations that make ChatGPT valuable. OpenAI has not supplied the evidence to answer that test yet.