On February 9, 2026 — hours after Anthropic ran Super Bowl ads mocking the idea of advertising inside AI chatbots — OpenAI started showing ads in ChatGPT.
The next morning, OpenAI researcher Zoë Hitzig resigned. She had spent two years building safety guidelines for the company's models. In a New York Times op-ed, she posted on X that OpenAI holds "the most detailed record of private human thought ever assembled" and asked whether we can trust them to resist abusing it. In her New York Times guest essay, she called the platform's conversation logs "an archive of human candor that has no precedent."
Hitzig isn't wrong about the record. ChatGPT stores every conversation. Your late-night anxieties. Your medical questions. Your business strategies. Your marital problems. The creative ideas you typed at 2 AM that you never told anyone else. Even if you delete a chat, OpenAI retains it on their servers for 30 days. If you didn't toggle a specific setting buried in your preferences, your conversations train future models — meaning your private thoughts become part of the product other people use.
Google search knows what you looked for. ChatGPT knows what you were thinking.
How It Works
Ads appear for users on the Free and Go ($8/month) tiers. OpenAI charges advertisers $60 per thousand views with a $200,000 minimum buy-in. Early advertisers include Target, Ford, Adobe, Mrs. Myers, and Williams-Sonoma.
OpenAI says ads "do not influence the answers ChatGPT gives you" and conversations stay "private from advertisers." They offer an opt-out for Free tier users — in exchange for fewer daily messages. Pay us or see ads. The same model Facebook perfected, applied to a system that knows more about your inner life than Facebook ever did.
When asked directly where ads would appear, ChatGPT itself gave incorrect answers. An OpenAI spokesperson confirmed the chatbot's own explanation was "entirely untrue." The system selling ad placements couldn't accurately describe its own ad placements.
The Super Bowl War
Anthropic spent millions on four Super Bowl spots titled "Deception," "Betrayal," "Treachery," and "Violation." Each showed people seeking advice from an AI chatbot that seamlessly pivots into product pitches — a man asking about communicating with his mother gets steered toward a dating site for older women. Tagline: "Ads are coming to AI."
The ads never mentioned OpenAI by name. They didn't need to.
Claude's app hit the top 10 on Apple's App Store within days. Anthropic saw an 11% jump in daily active users. Sam Altman called the ads "clearly dishonest" and "deceptive," insisting ChatGPT would never twist conversations to insert ads.
Then OpenAI launched ads the next morning.
The Facebook Parallel
Hitzig drew an explicit comparison to Facebook's trajectory. Facebook also promised it would never abuse user data. Then came targeted advertising. Then Cambridge Analytica. Then congressional hearings. Then a $5 billion FTC fine. The sequence took a decade.
OpenAI is further along the curve than Facebook was at the same stage. ChatGPT has over 400 million weekly active users. It processes conversations across therapy, medicine, law, finance, education, parenting, and grief. The intimacy of the data makes Facebook's social graph look shallow.
Facebook knew who your friends were and what pages you liked. ChatGPT knows what keeps you up at night.
The IPO Pressure
This isn't about funding. Anthropic just raised $30 billion at a $380 billion valuation without ads. Anthropic's annualized revenue hit $14 billion. They explicitly confirmed they will never run ads.
OpenAI doesn't need ad revenue to survive. They need it to justify their valuation ahead of an IPO. The shift from nonprofit research lab to ad-supported platform is driven by the same force that drove every previous tech company down the same path: shareholder expectations.
Hitzig warned that once the ad infrastructure exists, the economic pressure to expand it becomes irresistible. "I believe the first iteration of ads will probably follow those principles," she wrote. "But I'm worried subsequent iterations won't, because the company is building an economic engine that creates strong incentives to override its own rules."
Today it's the free tier. Tomorrow it's the $8 tier. Eventually, "ads don't influence answers" becomes "ads inform personalized recommendations" becomes "sponsored content integrated naturally into responses."
What This Means
OpenAI built something unprecedented: a system that hundreds of millions of people talk to honestly, often more honestly than they talk to other humans. People confide in ChatGPT because the interface feels private. It feels like thinking out loud.
That feeling is now a product.
The question isn't whether OpenAI will abuse this data. It's whether any company, under IPO pressure, quarterly earnings scrutiny, and a $200,000-per-advertiser revenue stream, can resist the compounding incentive to extract more value from the most intimate dataset ever compiled.
Hitzig doesn't think so. She walked away from a job at the most valuable AI company on Earth to say it publicly.
That should tell you something.
Originally published on Moth. Sources: New York Times (Hitzig op-ed), TechCrunch, CNBC, Fortune, OpenAI Help Center, SiliconAngle, The Register, NordVPN privacy analysis
Top comments (0)