OpenAI is making the same mistake Facebook did, researcher says as she quits over ChatGPT ads

OpenAI

• Zoë Hitzig resigned from OpenAI over concerns about ads in ChatGPT
• She warns ads could risk user trust and privacy, comparing OpenAI to Facebook
• OpenAI says ads won't influence answers or appear in sensitive conversations

Did our AI summary help?

A former OpenAI researcher has resigned, warning that the company may be heading down the same path Facebook once did when it embraced advertising.

Zoë Hitzig, an economist and poet who spent two years at OpenAI, announced in a guest essay in The New York Times that she stepped down earlier this week. Her resignation came the same day OpenAI began testing ads inside ChatGPT.

Hitzig said the decision to introduce advertising raised serious concerns for her. She did not say that ads are wrong in themselves. Instead, she argued that ChatGPT is different from social media platforms because of the deeply personal things users share with it.

According to her, people have opened up to ChatGPT about medical worries, relationship troubles, and even religious beliefs. Many users felt comfortable doing so because they believed they were speaking to a tool with no hidden agenda. She described this data as “an archive of human candor” unlike anything seen before.

That, she warned, is what makes advertising risky.

Hitzig drew a clear comparison to Facebook, which in its early days promised users more control over their data. Over time, she wrote, those promises weakened. Regulators later found that some privacy changes presented as empowering users actually reduced their control.

She fears something similar could happen at OpenAI. While the company says ads will be clearly labeled and will not influence ChatGPT’s answers, Hitzig worries that once advertising becomes part of the business model, financial pressure could slowly change how decisions are made.

OpenAI recently said it would test ads for users on its free and lower-cost plans in the United States. Paid subscribers on higher tiers will not see ads. The company has said advertisers will not get access to personal chat data and that ads will not appear next to conversations about health, mental health, or politics.

Still, personalization is turned on by default in the test, meaning ads may be selected based on users’ past conversations and activity.

The debate over ads in AI has also sparked tension in the industry. Rival company Anthropic publicly said its chatbot Claude would remain ad-free and even ran ads suggesting that AI with product placements could feel intrusive. OpenAI CEO Sam Altman dismissed those claims as exaggerated and said advertising could help make AI accessible to people who cannot afford subscriptions.

Hitzig, however, believes the issue goes deeper than just ads versus no ads. She suggested alternative models, such as having business users subsidize free access for others or creating independent oversight boards to control how user data is used.

She ended her essay with a warning: the real danger, she said, is building a system that either manipulates users without them realizing it, or one that only serves those who can afford to pay.

Her resignation comes at a time when several AI researchers across different companies are stepping down, as the industry moves quickly from research labs to profit-driven products.