Campbell Brown, who spent six years as Meta's head of news partnerships before departing in 2023, says the AI industry has a gatekeeping problem that nobody is talking about. In a TechCrunch interview, Brown warned that the companies building AI models are making editorial decisions about what information reaches billions of people — without any of the transparency, accountability, or professional standards that traditional media institutions developed over centuries.
The Core Argument
Brown's thesis is simple. When you ask an AI chatbot a question, someone decided what data the model was trained on. Someone decided how the model weights certain sources over others. Someone decided what the model refuses to say. Those decisions shape what hundreds of millions of people believe to be true. And unlike a newspaper editor or a TV news director, the people making those decisions are not accountable to anyone.
At Meta, Brown saw firsthand how algorithmic decisions about news distribution affected public discourse. Facebook's News Feed shaped what billions of people read. The decisions about which publishers to amplify and which to suppress had enormous consequences. She argues AI is repeating the same pattern — but with even less transparency.
When Google AI search pulls answers from Reddit threads, someone decided Reddit is a trustworthy source. When ChatGPT refuses to discuss certain topics, someone wrote the policy. When Claude provides health advice that is wrong half the time, the training decisions that produced those errors were made behind closed doors.
Why It Matters Now
The timing of Brown's warning is significant. AI is no longer a novelty. Over 400 million people use ChatGPT. Google Gemini is embedded across Chrome, Maps, Workspace, YouTube, Google TV, and millions of vehicles. Meta AI handles 10 million business conversations per week. And Microsoft Copilot has 20 million paid enterprise users.
These tools are becoming primary information sources. For many users, asking an AI is replacing searching the web. The AI's response is treated as authoritative — even when it is incomplete, biased, or fabricated. The Harvard study showed AI can outperform doctors at diagnosis. The BMJ study showed it gives bad health advice half the time. Both are true simultaneously. Which version a user encounters depends on decisions made by people they will never meet.
The Accountability Gap
Brown pointed to a fundamental asymmetry. Newspapers have editors who are publicly accountable. TV news has anchors whose credibility is on the line. Social media platforms — after years of pressure — developed content moderation policies, transparency reports, and oversight boards.
AI companies have none of that. Training data decisions are proprietary. Safety policies are internal. And when an AI produces harmful output — whether fabricated medical credentials, misleading health advice, or AI writing patterns that masquerade as human content — there is no editor to fire, no correction to publish, and no accountability mechanism that the public can access.
OpenAI's own policy paper acknowledged it cannot manage AI risks alone. But calling for government involvement is different from accepting external oversight. Brown argues the AI industry wants to set its own rules — just as social media companies did before governments forced transparency.
What She Proposes
Brown stopped short of specific policy recommendations. But she argued for three principles. First, transparency about training data and editorial decisions. Users should know what sources an AI draws from and what it was told to avoid. Second, independent auditing. Third-party reviewers should evaluate AI outputs the way fact-checkers evaluate news articles. Third, professional standards. The AI industry needs something equivalent to journalism's ethical codes — guidelines about accuracy, fairness, and accountability that carry real consequences.
Whether the AI industry adopts these standards voluntarily or waits for regulation is, in Brown's view, the defining question. Social media waited. The result was years of misinformation, election interference, and public trust erosion before governments intervened. AI is on the same trajectory — but moving faster.
The Bigger Picture
Brown's warning connects to a growing unease about who controls AI's influence on public discourse. Barry Diller said trust in AI leaders is irrelevant because even they do not know what their technology will do. The Musk vs Altman trial is fundamentally about governance. And Anthropic's research showed that fictional portrayals of evil AI directly shaped how its model behaved.
The question of who decides what AI tells you is not abstract. It affects every person who asks an AI a question. And right now, the answer is a small group of engineers and executives at a handful of companies — making decisions that shape what billions of people believe, without any obligation to explain why.
