OpenAI began rolling out advertisements in ChatGPT this week, and the move has already cost them a key researcher. Zoe Hitzig, who spent two years shaping safety policies and AI model development at the company, resigned on Monday in protest. Her departure and the concerns she raised deserve serious attention from anyone building or deploying AI systems.
The advertising rollout targets users on the free tier and the newer $8 per month Go subscription plan. Plus, Pro, Business, Enterprise, and Education subscribers will not see ads. OpenAI positions this as expanding access to AI for those who cannot afford premium subscriptions. But the implementation raises questions that go beyond simple business model decisions.
The Archive of Human Candor
In her New York Times essay titled "OpenAI Is Making the Mistakes Facebook Made. I Quit," Hitzig highlighted something that should concern every AI practitioner. Users share deeply personal information with ChatGPT: medical concerns, relationship problems, religious questions, career anxieties. They do this because they perceive the AI as a neutral tool with no ulterior agenda.
As Hitzig put it, ChatGPT users have created an "archive of human candor that has no precedent." People confide in ChatGPT in ways they might not confide in search engines or social media platforms. The introduction of advertising changes the fundamental relationship between user and system.
OpenAI states that ads appear in a clearly labeled section beneath the chat interface, not inside responses. The company also claims ads do not influence the answers ChatGPT provides. But the economic incentives now point in a concerning direction.
How the Ad System Works
The technical implementation shows thoughtful design in some areas. Ads match conversation topics, past chats, and prior ad interactions. A user researching recipes might see ads for meal kits or grocery delivery. The system excludes sensitive topics including health, mental health, and politics from ad placement.
Users get granular controls: they can dismiss ads, view and delete ad history and interest data, and toggle personalization on or off. This is more control than most ad-supported platforms offer.
But these safeguards exist at launch. Hitzig expressed concern that "subsequent iterations" might not maintain these ethical boundaries, because OpenAI is now building "an economic engine that creates strong incentives to override its own rules."
This is the core tension. Once advertising revenue becomes a significant part of the business model, the pressure to maximize that revenue only grows. Facebook's evolution from connecting friends to optimizing engagement at any cost is the cautionary tale.
The Business Model Problem
OpenAI's shift to advertising reveals the fundamental challenge facing frontier AI companies. Training and running models at this scale requires massive capital. OpenAI reportedly spent over $7 billion on compute in 2025 alone. Subscription revenue, while growing, may not cover these costs at the current pricing structure.
The company is also preparing for a potential IPO, which would bring shareholder pressure for revenue growth. Advertising is a proven path to scaling revenue quickly. Google, Meta, and most of the internet run on this model.
But AI assistants are different from search engines or social media feeds. The conversational interface creates a sense of intimacy and trust that other platforms do not have. Monetizing that trust carries risks that may not be immediately apparent in quarterly revenue reports.
What This Means for Enterprise AI
For organizations deploying AI, particularly in regulated industries or government contexts, this shift matters. If you are using ChatGPT for any business purpose, understanding the data practices around advertising is now essential.
The enterprise tiers remain ad-free, but the broader direction should inform vendor selection. When evaluating AI providers, consider not just current capabilities but the business model pressures that will shape future development.
In the UAE and across the Middle East, where data sovereignty and user privacy carry significant weight in procurement decisions, the advertising model introduces new considerations. Organizations may prefer AI providers with business models that do not depend on monetizing user interactions.
The Competitive Response
Anthropic, OpenAI's primary competitor, ran a Super Bowl ad this year that specifically mocked the idea of AI platforms including advertising. The ad highlighted Claude as an alternative that prioritizes user interests over ad revenue. Whether this differentiation remains sustainable long term is an open question, but for now it represents a clear market positioning.
For AI practitioners, this competitive dynamic is worth watching. The business models that win in the current phase of AI development will shape the industry for years to come.
Looking Forward
OpenAI's advertising launch marks a significant moment in the commercialization of AI assistants. The immediate impact may be minimal for most users. But the structural incentives are now in place for the kind of gradual erosion of user interests that defined the social media era.
Hitzig's resignation is a signal that should not be dismissed. When researchers with deep knowledge of a company's inner workings leave over ethical concerns, their warnings deserve serious consideration. The next few years will reveal whether OpenAI can resist the pressures that advertising revenue creates, or whether ChatGPT follows the path that Facebook walked before it.
For those of us building AI systems and advising organizations on AI adoption, this is a reminder that technical capabilities are only part of the evaluation. Business models, incentive structures, and corporate governance matter just as much when choosing AI partners for the long term.