Campbell Brown: Artificial intelligence is quickly becoming one of the main ways people search for information. Instead of opening multiple websites, many users now ask AI chatbots for answers about news, health, finance, politics, work, and everyday decisions.
That raises a major question: who decides what AI tells you?
Campbell Brown, the former news chief at Meta, believes this question is becoming more urgent. After years of working in journalism and social media, she now runs Forum AI, a company focused on evaluating how AI models respond to complex and high-stakes topics.
Her concern is simple: AI is becoming the new information gateway, but many systems are still not reliable enough for that role.
Campbell Brown New Focus: AI Accuracy
Campbell Brown has spent much of her career around information. She worked as a TV journalist and later became Facebook’s first dedicated news chief. Now, she is focused on how AI systems handle difficult questions.
Her company, Forum AI, evaluates foundation models on topics where the answers are not always simple. These include areas like:
- Geopolitics
- Mental health
- Finance
- Hiring
- Public policy
- Other high-stakes subjects
These are topics where a wrong or biased answer can create real-world problems. Brown’s argument is that AI companies should not only focus on coding, math, and technical benchmarks. They also need to focus on whether their systems give accurate, balanced, and useful information to regular people.
Why High-Stakes AI Answers Matter
AI tools are no longer just used for fun prompts or simple writing help. People are using them to understand serious topics.
A student may ask AI about world politics. A job seeker may use it for career advice. A small business owner may ask about finance. A person going through a difficult moment may ask about mental health.
If the answer is incomplete, biased, misleading, or missing important context, the user may not even realize it.
That is why Brown believes AI systems need better evaluation. The goal should not be only to sound confident. The goal should be to get closer to what is true, balanced, and useful.
How Forum AI Tests AI Models
Forum AI works by bringing in experts to help design better benchmarks for AI models.
For example, in geopolitics, Brown has worked with well-known experts and public figures, including Niall Ferguson, Fareed Zakaria, Tony Blinken, Kevin McCarthy, and Anne Neuberger.
The idea is to have experts create strong evaluation standards. Then, AI judges are trained to test models at scale. According to Brown, Forum AI aims for AI judges to reach about 90% consensus with human experts.
This approach is different from simple checklist-style testing. Instead of only checking whether an AI model passes basic prompts, Forum AI looks at how models handle nuance, context, missing perspectives, and edge cases.
The Problem With AI Answers Today
Brown says many AI models still make serious mistakes when answering information-based questions.
Some problems include:
- Missing important context
- Giving one-sided answers
- Leaving out key perspectives
- Creating weak or misleading arguments
- Pulling from questionable sources
- Showing political bias
- Producing answers that sound confident but are not fully accurate
This is one of the biggest challenges with AI. A chatbot can sound polished and helpful even when its answer is incomplete or wrong.
For average users, that makes it harder to know when to trust the response.
The Social Media Lesson
Brown’s experience at Facebook shaped how she thinks about AI.
She saw how social media platforms struggled with news, misinformation, and engagement-driven systems. Platforms often rewarded content that kept people clicking, reacting, and sharing, even when that content did not make people better informed.
Brown worries AI could repeat the same mistake.
If AI companies optimize mainly for what users want to hear, the results could become harmful. But if they optimize for accuracy, truthfulness, and better information, AI could become a healthier information system than social media ever was.
Enterprise AI Could Push for Better Accuracy
One reason Brown is hopeful is the enterprise market.
Businesses using AI for important decisions cannot afford unreliable answers. Companies working in areas like credit, insurance, lending, hiring, and compliance need AI systems that are accurate and legally safer.
If an AI tool gives a wrong answer in these fields, it can create liability. That means businesses may push AI companies to build models that are better evaluated, more reliable, and less biased.
Brown believes this business pressure could help move AI toward accuracy instead of engagement.
Also read;
Apple iOS 26.5 Update- Latest Features, Release Date and How to Install on Your iPhone
Google Fitbit Air 2026: Amazing Features, Price, Specs and Release Date
PlayStation Store 2026: How It Works, Sales, Dynamic Pricing and Best Ways to Save Money
Compliance Testing May Not Be Enough
Brown also argues that current AI compliance testing is often too weak.
Many audits rely on standard benchmarks or checkbox-style reviews. But high-stakes AI use requires deeper testing with real subject experts.
For example, hiring AI tools may pass a basic bias test but still fail in more complicated real-world situations. According to Brown, true evaluation needs domain expertise, edge-case testing, and careful review of how models behave in messy situations.
In simple words, smart general testing is not enough. AI systems need expert-level evaluation when the topic is serious.
Why Regular Users Still Don’t Fully Trust AI
There is a gap between how Silicon Valley talks about AI and how many everyday users experience it.
Tech leaders often describe AI as world-changing technology that will transform work, education, science, and business. But regular users often still see AI giving wrong answers, shallow summaries, or confusing responses.
That gap matters. If users do not trust AI, they may avoid it. But if they trust it too much without verification, they may be misled.
The real challenge is building AI that deserves trust.
Why This Debate Matters
The question of who controls AI answers is becoming one of the biggest issues in technology.
AI systems are starting to shape how people learn, search, work, and make decisions. If these systems become the main way people access information, then accuracy, bias, transparency, and expert review become extremely important.
Campbell Brown’s argument is that AI companies must take this responsibility seriously before the problems become too big to fix.
Conclusion
Campbell Brown’s work with Forum AI highlights a major concern in the AI industry: AI tools are becoming information gatekeepers, but they are not always accurate, balanced, or reliable.
Her message is not that AI should be avoided. Instead, she believes AI needs better testing, stronger expert review, and more focus on truth rather than engagement.
As more people rely on AI for serious questions, the future of information may depend on whether these systems are built to give users what is easy, or what is actually right.
Who is Campbell Brown?
Campbell Brown is a former TV journalist and former Meta news executive. She later founded Forum AI, a company focused on evaluating how AI models handle complex and high-stakes information.
What is Forum AI?
Forum AI is a company that evaluates foundation AI models on serious topics such as geopolitics, mental health, finance, hiring, and public policy.
Why is Campbell Brown concerned about AI?
She is concerned that AI chatbots are becoming a major source of information, but many models still give incomplete, biased, or inaccurate answers.
What are high-stakes AI topics?
High-stakes AI topics are areas where wrong information can cause real harm. These include mental health, finance, hiring, geopolitics, law, insurance, lending, and public policy.
How does Forum AI test AI models?
Forum AI works with domain experts to create benchmarks, then uses AI judges to evaluate model responses at scale. The goal is to measure whether models give accurate, balanced, and expert-aligned answers.
Why is AI accuracy important?
AI accuracy is important because more people now use chatbots to learn, make decisions, and understand serious issues. Wrong or biased answers can mislead users.
Can AI replace search engines?
AI is already changing how people search for information, but it still has problems with accuracy, context, and trust. Many users still need to verify AI answers with reliable sources.
What is AI bias?
AI bias happens when an AI system gives responses that unfairly favor one viewpoint, group, source, or interpretation over another.
Why does enterprise AI care about accuracy?
Businesses using AI for hiring, lending, insurance, finance, and compliance face legal and financial risks if AI systems give wrong or biased answers.
What is the biggest challenge for AI information tools?
The biggest challenge is making AI systems that are not only fast and helpful, but also accurate, balanced, transparent, and trustworthy.
