“`html
AI Has Editors and Campbell Brown Is Naming Them
Campbell Brown ran editorial decisions at Meta that reached billions of people. Now she’s pointing at the same power structure inside AI systems. A small group of people decide what AI tells you, how it frames answers, and what it leaves out. According to Humanity AI, more than $18 million in new grants landed in May 2026 specifically to study this problem. That’s not a coincidence. That’s a fire alarm.
Why This Is Happening Right Now
Brown spent years as Meta’s head of news partnerships. That’s not a communications role. That’s an editorial role. She decided which publishers got amplified, which stories got surfaced, and which signals the algorithm rewarded. She was, functionally, one of the most powerful editors in the history of media. Most people never heard her name.
Now she’s talking openly about the exact same structure sitting inside AI. And the timing tracks. According to Humanity AI, the organization announced more than $18 million in new grants in May 2026 to researchers at the AI Now Institute and DAIR to specifically study AI’s impact on labor and democratic institutions. That money went to groups that have always asked the same question Brown is asking: who decides?
On May 12, 2026, Celonis launched the Celonis Context Model and signed an agreement to acquire Ikigai Labs, according to Celonis Press Room. Their core argument was that AI agents fail because they operate inside a “context gap.” They act on incomplete pictures of reality and produce confident output anyway. Brown’s argument about AI content control is structurally identical. The user gets a confident answer. They don’t know what got left out to build it.
The Power Nobody Talks About
Here’s what I think gets missed in most coverage of this topic. People worry about AI hallucinating. That’s real but it’s the visible failure. The invisible failure is selective framing. AI doesn’t have to lie to mislead you. It just has to consistently frame one type of answer as reasonable and another type as fringe.
That framing comes from training data choices. It comes from alignment decisions. It comes from safety filter guidelines written by teams at AI companies who are employed by people who answer to boards and investors. None of that is disclosed when you type a question and get an answer.
Brown knows this because she built the earlier version of it. Facebook’s News Feed wasn’t random. Every amplification decision had logic behind it, and that logic was built by people with perspectives, regulatory pressures, and career incentives. The AI layer works the same way with less transparency and more users.
According to a May 2026 study by Akeyless, reported by Morningstar, 67% of enterprises already suspect AI agents have accessed data beyond their intended scope, with detection taking an average of 14 hours. That’s in controlled corporate environments with dedicated IT teams. Now apply that detection failure to information systems used by people who don’t know there’s anything to detect.
The Humanity AI grants, according to Ford Foundation reporting, specifically went to the AI Now Institute and DAIR. These are not neutral research bodies. They’re organizations that have built their reputations on asking hard structural questions about AI power. Who owns the model? Who can change what it says? Who profits when it nudges users toward one political or commercial conclusion over another? Those are the questions Brown is raising. The fact that $18 million is now chasing those same questions tells you how serious the power gap actually is.
I’ve watched this pattern run three times in my lifetime. Print editors shaped public opinion for a century with no names attached to specific decisions. TV news anchors did it for fifty years. Social media CEOs did it for twenty. Every time, the public learned about the control structure after the damage was done. Every iteration said it was different. AI is not different. It’s just faster, harder to audit, and operating at a scale none of the earlier versions reached.
If you’re trying to make sense of AI governance stories and reach an audience with them, tools like InVideo AI let you turn complex structural arguments into short video content that actually gets watched. The editorial power debate is visual and urgent. It’s also being lost right now because the people who understand it aren’t producing content fast enough to compete with the people who benefit from confusion.
What This Means For You
Here’s what I would do with this information.
Stop treating AI answers as neutral. Every time an AI system gives you information about politics, regulation, health policy, or financial markets, ask the same question you’d ask about any other media source: who built this, what did they leave out, and what do they benefit from my believing this version?
Follow the funding trail on AI oversight research. According to Humanity AI’s May 2026 announcement, $18 million went to independent researchers studying AI’s democratic impact. That kind of funding signals where the contested terrain actually is. Follow it and you’ll find the real arguments before they become mainstream stories.
Push for named accountability on AI editorial decisions. Anonymous editorial decisions are not acceptable at the scale AI operates. If a news editor at a major publication makes a call that affects ten million readers, their name is attached to it somewhere in the chain. AI editorial decisions affect far more people with no names attached to anything. That gap is not a technical limitation. It’s a choice.
Watch what enterprise AI control failures tell you about consumer AI. According to Celonis Press Room, the Context Model they launched on May 12, 2026 is built specifically to give AI agents an accurate operational picture before they act. The principle applies directly to information AI. Context determines output. Whoever controls the context controls the answer.
If you’re building content around AI governance or any fast-moving tech story, AppSumo has lifetime deals on research and writing tools worth looking at. This story is moving quickly and production speed matters when you’re trying to stay ahead of the news cycle rather than react to it.
The Bottom Line
Campbell Brown spent years making the editorial machine work. Now she’s explaining how it works. Most people still don’t believe it works the way it does. AI doesn’t have opinions, but the people who build AI do. Right now, those people are making daily decisions that shape what billions of humans believe about the world, about money, about politics, and about each other. That’s not neutral infrastructure. That’s the most concentrated editorial power in human history sitting behind a chat box. The people with access to it know exactly what they have. The rest of us are just catching up.
Frequently Asked Questions
Who is Campbell Brown and why does her take on AI content control matter?
Campbell Brown served as Meta’s head of news partnerships, one of the most operationally powerful editorial roles in modern media. She made real decisions about what content billions of people saw, at scale, for years. Her perspective on AI content control carries weight because she’s describing a system she helped build from the inside, not speculating about one she’s read about.
How do AI systems actually decide what information to surface or frame?
AI systems are trained on data sets chosen by humans, filtered through safety guidelines written by humans, and shaped by alignment decisions made by teams working at companies with investors and boards. According to Celonis Press Room, the biggest structural failure in enterprise AI is the “context gap,” where the system acts on an incomplete picture of reality. The same gap exists in information AI, where missing context and omitted framing shape every answer the system delivers.
Is there independent oversight of what AI tells users?
Not in any meaningful regulatory form as of 2026. According to Humanity AI’s May 2026 announcement, more than $18 million in new grants went to groups like the AI Now Institute to study AI’s impact on democratic institutions. That research is early and independent. Formal regulatory oversight of AI content decisions barely exists in most jurisdictions and enforcement is ly nonexistent.
Why should I care about AI editorial decisions if I’m not in politics or media?
Because AI editorial decisions shape financial information, health information, and legal information just as much as political content. According to a Morningstar report citing Akeyless research, 67% of enterprises already suspect AI agents have accessed data beyond their intended scope. The same audit failures that allow unauthorized data access also allow undisclosed editorial shaping in consumer tools. The structural problems are identical.
What can I do to reduce AI editorial bias in the information I consume?
Check AI answers against primary sources on any topic with political or financial stakes. Follow funding disclosures at major AI safety and oversight organizations to understand who is paying for which research conclusions. And demand named accountability on AI content decisions the same way journalism standards demand it from human editors. Anonymity at this scale isn’t a technical constraint. It’s a policy choice that benefits the people making the decisions.
“`