OpenAI builds offerings on consent as it expands ChatGPT ads to Europe

OpenAI’s hand is mostly forced: under the General Data Protection Regulation (GDPR), personalized ads count as processing personal data and OpenAI needs a legal basis to run them. Some ad platforms rely on “legitimate interest” to justify targeting without asking permission first. OpenAI is being more explicit and will ask users for consent — which could (theoretically) enable OpenAI to serve more targeted ads.

The move marks the next major stage of OpenAI’s ad rollout. Since its initial launch in the U.S. in February, the ad pilot is currently available in nine markets, including Canada, Australia, New Zealand, the U.K. Japan, South Korea, Mexico and Brazil. OpenAI had already laid the groundwork for ads in Europe in its June 4 privacy policy, which stated the company could use data from free and Go users to personalize ads, but did not offer a timeline.

What could be ahead for ChatGPT ads in Europe

OpenAI could feasibly offer users to choose between personalized ads (ads served using information about the user) and contextual ads (ads served without reference to personal data), based on the user’s conversation, according to the outlined policy.

For users who are using free accounts and Go ChatGPT, its lower-cost subscription tier, OpenAI can use a range of information to make the ads they see more relevant. The company can use personal data to “personalize the ads you see” and measure how well those ads perform. That can include information from advertisers and other partners, such as what someone bought after clicking an ad, as well as a user’s “ads history and interest” — including the types of ad content they actually view or click on. Free and Go users can choose what data OpenAI uses for that personalization through “advertising controls in account settings”, according to the policy.

Users who don’t consent to personalized ads will still see ads. Instead, OpenAI will use contextual information to decide which ads to show them, with targeting narrowed to the live conversation and a rough location while chat history and memory are left out. The policy states users have the right to “withdraw your consent”, given that the company relies “on consent as the legal basis for processing”.

That distinction could leave some users feeling like they have to pay to avoid ads altogether, said Clara Westbrook, head of data privacy at Arbor Law.

“I think the fact it only applies to users on the more basic plans, which are free, could effectively result in users feeling pressured to go onto a paid plan to avoid ads if they do not wish their data to be used for the service of ads, personalised [sic] or otherwise,” she said. However, the breakdown of users per Free, Go and paid tiers is not publicly available, The Information reported that around 35 million of ChatGPT’s 900 million-strong global user base were paying for Plus or Pro subscriptions as of last month.

“It almost feels like people are having to pay if they don’t want to receive ads at all,” Westbrook added.

But OpenAI is doing something most platforms were forced into doing as the law went into effect in 2018. 

Meta, for instance, spent years fighting for the right to run behavioral ads without asking for permission first. It claimed the ads were part of its contract with users. Regulators rejected that and fined Meta €390 million ($451.9 million) in January 2023. Meta then switched to a legal basis called “legitimate interest”, essentially arguing it had a good enough reason to use people’s data that it didn’t need to ask. The European Data Protection Board subsequently determined that Meta could not rely on legitimate interests for processing, following a July 2023 ruling by the EU’s top court that addressed Meta’s legal basis for processing. Norway banned the practice outright weeks later, and EU regulators extended that ban across the whole block that October. Consent became Meta’s legal basis because every other option had been shut down.

OpenAI declined to comment on this article.