OpenAI researcher Zoë Hitzig resigns over ChatGPT advertising plans
OpenAI researcher Zoë Hitzig resigned on 12 February 2026 after the company began testing advertisements on ChatGPT. She warned that because users share deeply personal information with the chatbot based on trust, data-driven advertising could enable manipulation, and that revenue dependence on ads could eventually pressure the company to relax safeguards against targeting vulnerable users.
What happened
Researcher Zoë Hitzig resigned from OpenAI on 12 February 2026 after the company began testing advertisements on ChatGPT.
Hitzig had spent two years at OpenAI helping to define how to create and price AI models and framing initial safety policies. In a New York Times essay, she expressed concerns that OpenAI had stopped asking the questions she joined to answer.
Reported impact
Hitzig warned that ChatGPT users share personal information including health concerns, relationship issues and religious beliefs with the chatbot, creating an archive of human sincerity based on trust that the AI has no hidden agenda. She cautioned that data-driven advertising could exploit users' hidden fears and desires, creating significant risks of manipulation.
Hitzig expressed doubts about OpenAI's advertising strategy, stating: 'I don't think ads are immoral or unethical. AI is expensive to operate, and ads can be an essential revenue source. But I have doubts about OpenAI's strategy.'
Index analysis
Hitzig argued the financial dilemma is fundamentally a moral one: either restrict technology access to a financially privileged segment, or let ads target users' fundamental fears to sell them products. She stated: 'This is a false choice. Technology can and must remain accessible at scale, while simultaneously monitoring, profiling and preventing manipulation of users.'
Hitzig warned that even if initial policies prevent ads from influencing response content, once the business model becomes dependent on advertising revenue, internal pressure will emerge to relax those policies. She noted that social media advertising similarly evolved gradually toward sophisticated targeting and persuasion.
Governance controls
OpenAI stated it will display ads clearly and that they will not influence answer content. However, Hitzig cautioned that dependence on advertising revenue could create pressure within the company to weaken such safeguards over time.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.