A former OpenAI researcher, Zoe Hitzig, resigned and publicly criticized the company for prioritizing profit over its original mission of ethical AI development, particularly highlighting concerns about the introduction of ads in ChatGPT and the potential for user manipulation. The video warns that without strong oversight or alternative business models, OpenAI risks repeating the harmful patterns of other tech giants, eroding user trust and privacy for financial gain.
A former OpenAI researcher, Zoe Hitzig, recently quit the company and published a guest essay in The New York Times, raising concerns about OpenAI’s shift in priorities. Hitzig, who spent two years at OpenAI shaping AI model development and pricing, resigned on the same day the company launched advertisements in the free version of ChatGPT. She argues that OpenAI has moved away from its original mission of building AI safely and ethically, instead prioritizing profit over user well-being and safety. This shift is particularly troubling given OpenAI’s roots as a nonprofit research organization, now transformed into a for-profit entity with massive financial ambitions.
Hitzig’s main concern centers on the introduction of ads into ChatGPT, which she believes fundamentally changes the user-AI relationship. Previously, users interacted with ChatGPT under the assumption that it had no ulterior motives, often sharing private thoughts and seeking advice. The introduction of ads, especially in such an intimate digital space, creates incentives for OpenAI to manipulate users for profit, potentially exploiting their vulnerabilities. Hitzig warns that this could lead to the same gradual erosion of trust and privacy seen with companies like Facebook, where initial promises of user control gave way to relentless monetization and data exploitation.
The video also highlights a broader trend of researchers leaving OpenAI due to similar concerns. For example, another prominent alignment researcher left for Anthropic after disagreements over OpenAI’s allocation of resources to AI safety. These departures suggest a growing internal conflict between the company’s original ethical commitments and its current profit-driven trajectory. The video draws parallels to social media’s negative societal impacts, warning that optimizing AI for engagement and ad revenue could have even more profound consequences, such as encouraging unhealthy user dependence and even psychological issues like “LLM psychosis.”
To address these issues, Hitzig proposes several alternative business models and safeguards. One suggestion is to have large businesses that benefit most from AI pay surcharges to subsidize free access for regular users, similar to how utilities are funded. Another is to establish independent oversight boards with real authority to enforce ethical standards and protect user data, rather than relying on unenforceable company blog posts. A third idea is to create data trusts—independent organizations that control user data and grant access only with explicit permission, thereby flipping the power dynamic away from the company.
Despite these potential solutions, the video expresses skepticism that OpenAI or similar companies will voluntarily adopt them, given the immense financial incentives at stake. The likely outcome, the video argues, is that OpenAI will follow the well-worn path of other tech giants: introducing ads, gradually eroding user protections, and treating regulatory fines as a cost of doing business. Without strong external oversight or regulation, the company’s incentives will continue to favor profit over user welfare, potentially leading to a more manipulative and dystopian AI landscape.