On April 30, OpenAI emailed users a privacy policy update that most people probably skimmed past. Here is the part that matters: the company now shares your cookie IDs, device IDs, and email addresses with advertising partners to promote ChatGPT and Codex on third-party platforms like Instagram. If you are one of the 200 million-plus people using ChatGPT for free, this was turned on by default.

WIRED tested it. Two free accounts: tracking on by default. Two paying accounts (one Plus, one Enterprise): off by default. The message is clear. If you pay, you are the customer. If you do not pay, you are the product.

What Changed in the Privacy Policy

The headline change is a rebrand. The "Vendors and Service Providers" section of the privacy policy is now called "Vendors, Service Providers, and Marketing Partners." That last category is new. OpenAI is no longer just sharing your data with companies that help run its infrastructure. It is sharing data with companies whose entire job is to target you with ads.

The old policy said data went to partners who "access, process, or store Personal Data only in the course of performing their duties to us." The new version adds a critical sentence: OpenAI also shares "limited information with select marketing partners who are not service providers in order to promote our products and services on third-party properties." That is corporate for: we are giving your identifiers to ad networks.

Adweek's reporting adds another layer. OpenAI is not just pushing data out. It is pulling data in. The company now "openly acknowledges receiving data from advertisers and their partners, including information about purchases users make, to measure ad effectiveness." Read that again. OpenAI is receiving your purchase history from its advertising partners to figure out whether their ads on ChatGPT are working.

The Opt-Out Exists. Good Luck Finding It.

You can turn this off. Settings, then Data Controls, then Marketing Privacy. OpenAI spokesperson Taya Christianson confirmed users can "opt out at any time in settings." The company also insists conversations are not shared with advertising partners. Your chats stay private. It is the metadata about you, not the content of your conversations, that flows to ad networks.

But opt-in-by-default is a deliberate design choice. The company knows most people will never find that toggle. That is the whole point. The same pattern Facebook used for a decade. The same pattern Google used with location tracking. The same pattern every ad-supported platform has ever used. Make the tracking the default, make the opt-out obscure, and count on user inertia.

The Nonprofit-to-Ad-Platform Pipeline

This is the arc in one sentence: OpenAI started as a nonprofit to benefit humanity, converted to a capped-profit, converted again to a full for-profit, started testing ads inside ChatGPT in February, and is now sharing user data with advertising partners while receiving purchase data back from them. In about 11 years, it went from "open AI for the good of all" to "we will use your cookie ID to see if our Instagram ad got you to sign up for Codex."

The timing is important. This lands while Elon Musk is literally in court arguing that OpenAI betrayed its nonprofit mission. The company's answer to that accusation is apparently: let us also build an ad platform while we are at it.

And the market dynamics make this inevitable. Over 90 percent of ChatGPT users are on the free tier. That is hundreds of millions of people generating costs and producing zero revenue. OpenAI is hemorrhaging money on inference. The $40 billion fundraise, the IPO chatter, the enterprise push: none of it changes the math that most users will never pay $20 a month. The ad model is not a side hustle. It is the business model for the majority of the user base.

What to Watch

Three things. First, whether this triggers regulatory attention from the FTC or state privacy regulators. California's CCPA and the EU's GDPR both have opinions about opt-in-by-default tracking. Second, how Anthropic and Google respond. Claude does not run ads. Gemini is testing them but has not formalized ad-partner data sharing at this level. If Anthropic makes a "we do not track you" marketing push, this becomes a competitive story, not just a privacy one. Third, whether ChatGPT uninstalls accelerate. Adweek noted that uninstalls are already rising. If users start leaving over this, OpenAI faces the worst of both worlds: fewer users AND an ad platform nobody trusts.

First reported by WIRED and Adweek.