OpenAI can now use data from advertisers for more than ads

The change

On September 10, 2026, OpenAI published a new version of its US privacy policy. ChatGPT has shown ads to Free and Go users since February, so ads themselves are not new. What changed is what the policy says about the data around them: what OpenAI collects about the ads you see, what it receives from advertisers, and what it can do with that data.

The same wording first appeared in OpenAI's European privacy policy in August, when ads came to Europe. The US version now matches it. Venpo flagged the change the next day, and the redline is on the public change page.

What changed

Four edits change what the policy allows:

• Advertiser data is now a stated fact. The old text said OpenAI "may receive" information from advertisers and other data partners. The new text says "We receive" it. The example stays the same: OpenAI could receive information about purchases you make from those advertisers.

• That data can be used more widely. Advertiser data used to be tied to measuring and improving ads. It is now used "for purposes described in this Policy," including improving "our Services," with ads as one example.

• Ad activity is now named as collected data. For Free and Go users, the policy adds that OpenAI collects "information about your ads history and interests," such as the ads you view or engage with.

• Browser promises were removed. The policy now refers to an in-app browser instead of the Atlas browser, and it deleted two lines: that Atlas incognito history "won't be saved after you end your session," and that Atlas users can delete their browsing history or use incognito mode.

From "measure ads" to "purposes described in this Policy"

The most important edit is also the shortest. Before September 10, the policy described advertiser data with a narrow purpose: helping OpenAI measure and improve the ads shown to Free and Go users. After September 10, the same data is used for the purposes described across the whole policy.

Those purposes are broad. The policy lists providing and maintaining the services, improving and developing them, conducting research, and personalizing the experience, alongside ads. In plain terms, information that advertisers and data partners send OpenAI, such as a record that you bought something from them, is no longer limited on paper to checking whether an ad worked. It sits in the same pool as the rest of OpenAI's data.

The policy does not say how often this happens or what OpenAI does with it in practice. It is the permission that got wider, and permissions are what a privacy policy is for.

What did not change

Several protections still stand. The policy still says OpenAI does not "sell" personal data. OpenAI's help page on ads says advertisers never receive your chats, chat history, memories, name, email, precise location, or IP address, and that Free and Go users can turn off ad personalization and delete their ads data.

Business customers are also outside this change. Plus, Pro, Business, Enterprise, and Education plans do not show ads, and the privacy policy says it does not apply to content OpenAI processes for customers of its business offerings, such as the API. On September 21, OpenAI also added a table for U.S. state privacy laws that lists its data categories, purposes, and recipients in one place, which makes the ad-related uses easier to see.

Why this matters

For most SaaS companies, the risk is not the company's ChatGPT Business workspace. It is the employee who opens a personal Free account to rewrite a customer email, summarize a contract, or debug code. That account now sits under a policy where the chat context can shape ads, ad activity is collected, and data from advertisers can be used for any purpose the policy describes.

It is the same shift we have been tracking across AI vendors, where each update moves a little more data into a little more use, as when Perplexity moved three protections out of its privacy policy and Meta rewrote its policy around agentic AI. Keeping track of which AI tools touch company data, and under which terms, is the same discipline as monitoring AI subprocessors.

Potential impact

For a SaaS company whose team uses ChatGPT, the update raises four practical questions:

• Does your AI policy say work must happen on a business plan, and do people know that Free and Go accounts show ads and run under different terms?

• Do you know how many employees use personal ChatGPT accounts for work tasks?

• For anyone who uses Free or Go personally, is ad personalization turned off in Settings, if they prefer it off?

• Did anyone rely on the Atlas incognito and history promises for browsing on shared devices?

Answering questions like these across every AI vendor is part of the hidden work of vendor legal changes.

How Venpo detected it

Venpo monitors OpenAI's legal pages as part of continuous vendor risk monitoring. On September 11, the day after the new version went live, it flagged the update, marked every inserted and deleted line, and separated the ad and browser edits from a set of Sora references and link fixes that changed nothing. The full redline is on the OpenAI change page, and every monitored OpenAI document is listed on the OpenAI vendor profile.

Business outcome

Teams that track OpenAI got a plain-English note the day after the change, with the old and new sentences side by side. That leaves time to update the internal AI policy, remind the team which accounts are approved for work, and point personal users to the ad settings. The alternative is finding out from a customer's security questionnaire that asks how employees use AI tools.

Key takeaway

OpenAI's September update did not add ads, but it changed what the ads come with: a stated flow of data from advertisers, a wider set of uses for it, and fewer browser privacy promises. The difference between "may receive" and "we receive" is two words long, which is why the only reliable way to catch it is to monitor the text itself.