OpenAI’s move to test ads inside ChatGPT has turned into a larger argument about trust, privacy, and the business model behind AI assistants. Former OpenAI researcher Zoë Hitzig made that debate public when she resigned from the company on Monday and explained her decision in a guest essay in The New York Times.
Hitzig, an economist and published poet who holds a junior fellowship at the Harvard Society of Fellows, spent two years at OpenAI helping shape how its AI models were built and priced. Her warning is direct: ChatGPT ads could push OpenAI toward the same kind of incentives she believes damaged Facebook’s early promises to users.
Why the resignation matters
Hitzig did not argue that advertising is automatically immoral. Her objection is tied to the setting in which the ads appear. ChatGPT is not a feed of public posts or a search results page; users often bring it private questions, anxieties, plans, and beliefs.
In her essay, Hitzig wrote that people have shared medical fears, relationship problems, and religious beliefs with the chatbot, often "because people believed they were talking to something that had no ulterior agenda." She described the resulting store of disclosures as "an archive of human candor that has no precedent."
That framing makes the ad question more serious. If people treat an AI assistant like a place for private reflection, then even clearly labeled advertising can change the relationship. The concern is not only whether an advertiser receives personal details. It is whether the assistant’s economic purpose starts shaping the experience over time.
"I once believed I could help the people building A.I. get ahead of the problems it would create," Hitzig wrote. "This week confirmed my slow realization that OpenAI seems to have stopped asking the questions I’d joined to help answer."
How OpenAI says the ad test works
OpenAI announced in January that it would begin testing ads in the US for users on its free and $8-per-month “Go” subscription tiers. Paid Plus, Pro, Business, Enterprise, and Education subscribers would not see ads.
The company said the ads would appear at the bottom of ChatGPT responses, be clearly labeled, and would not influence the chatbot’s answers. Those design choices are important because they separate the ad placement from the answer itself, at least in the first version of the test.
OpenAI’s support documentation also says ad personalization is enabled by default for users in the test. If it remains on, ads are selected using information from current and past chat threads, along with past ad interactions. OpenAI says advertisers do not receive users’ chats or personal details, and ads will not appear near conversations about health, mental health, or politics.
Hitzig’s concern is that initial rules may not be enough once the company depends on an ad engine. She wrote: "I believe the first iteration of ads will probably follow those principles. 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."
The Facebook comparison
Hitzig connected the issue to Facebook’s early history. According to her essay, Facebook once promised users control over their data and the ability to vote on policy changes. Over time, she wrote, those commitments weakened.
The source article notes that the Federal Trade Commission found that privacy changes Facebook marketed as giving users more control actually did the opposite. For Hitzig, the lesson is not that every ad business repeats the same path automatically. It is that incentives can make yesterday’s limits easier to loosen tomorrow.
That is why the ChatGPT ad debate is bigger than whether an ad appears at the bottom of a response. The core question is what happens when a product built around conversation also has a reason to increase engagement, retain attention, and personalize commercial messages.
Hitzig pointed to an existing tension in OpenAI’s principles. She noted that the company says it does not optimize for user activity solely to generate advertising revenue, while reporting has suggested that OpenAI already optimizes for daily active users, "likely by encouraging the model to be more flattering and sycophantic."
Industry rivals are drawing lines
The timing also matters because OpenAI’s ad test followed a week of public sparring with Anthropic. Anthropic declared Claude would remain ad-free and ran Super Bowl ads with the tagline "Ads are coming to AI. But not to Claude," showing AI chatbots awkwardly inserting product placements into personal conversations.
OpenAI CEO Sam Altman called those ads "funny" but "clearly dishonest," writing on X that OpenAI "would obviously never run ads in the way Anthropic depicts them." He presented the ad-supported model as a way to make AI available to users who cannot afford subscriptions, writing that "Anthropic serves an expensive product to rich people."
Anthropic responded as part of its own advertising campaign, saying that ads inside conversations with Claude "would be incompatible with what we want Claude to be: a genuinely helpful assistant for work and for deep thinking." The company also said more than 80 percent of its revenue comes from enterprise customers.
This exchange shows how AI companies are beginning to define themselves not only by model capability, but by business model. One side argues ads can support wider access. The other argues that ads inside conversations change what an assistant is.
Alternatives beyond ads or subscriptions
Hitzig did not frame the only choices as paid access or ad-supported access. She proposed structural alternatives intended to preserve broad availability while limiting the pressure to use conversational data for targeting.
Her suggestions included cross-subsidies modeled on the FCC’s universal service fund, in which businesses that pay for high-value AI labor would subsidize free access for others. She also proposed independent oversight boards with binding authority over how conversational data is used in ad targeting.
Another idea was data trusts or cooperatives that would let users retain control of their information. Hitzig pointed to the Swiss cooperative MIDATA and Germany’s co-determination laws as partial precedents.
The broader AI industry is also seeing visible departures. Mrinank Sharma, who led Anthropic’s Safeguards Research Team and co-authored a widely cited 2023 study on AI sycophancy, announced his departure in a letter warning that "the world is in peril." At xAI, co-founder Yuhuai “Tony” Wu resigned on Monday, followed the next day by fellow co-founder Jimmy Ba. According to TechCrunch, at least nine xAI employees, including the two co-founders, publicly announced departures over the past week, and six of the company’s 12 original co-founders have now left.
The source article describes these departures across OpenAI, Anthropic, and xAI as unrelated in their specifics. Still, they arrive during a period of rapid commercialization that is testing researchers at major AI labs. Hitzig ended her essay by naming the two outcomes she fears most: "a technology that manipulates the people who use it at no cost, and one that exclusively benefits the few who can afford to use it."