Campbell Brown spent years chasing facts on television. She later joined Facebook as its first dedicated news chief. Now she says the next battle over truth has shifted from social media feeds straight into AI chatbots — and AI model accuracy sits at the center of that fight.
Brown, now co-founder and CEO of Forum AI, told TechCrunch that major technology companies have poured too much energy into coding and math benchmarks. She says they have largely ignored answers that shape how people understand news, politics, finance, hiring, and health care.
That neglect, she argues, turns AI model accuracy into a business risk, a media problem, and a growing public trust crisis.
Forum AI builds a market around expert judgment
Forum AI evaluates how leading AI systems respond to high-stakes questions. The company focuses on topics that resist clean yes-or-no answers — geopolitics, mental health, finance, and hiring.
Brown told TechCrunch these are subjects where “there are no clear yes-or-no answers, where it’s murky and nuanced and complex.”
Brown co-founded Forum AI with Robbie Goldfarb, a former Meta AI trust and safety lead. Lerer Hippeau confirmed in October 2025 that Forum AI had joined its investment portfolio. The firm said Forum AI deploys senior experts to test responses for bias, missing context, and tone.
Forum AI says standard benchmarks fail to measure real-world risk in banking, hiring, health care, and national security. The company trains judgment systems on expert reasoning and targets more than 90% accuracy against expert consensus.
Brown said Forum AI’s geopolitics work has drawn names, including Niall Ferguson, Fareed Zakaria, Tony Blinken, Kevin McCarthy, and Anne Neuberger.
Brown traced the idea to a specific moment at Meta.
“I was at Meta when ChatGPT was first released publicly,” she said, “and I remember really shortly after realizing this is going to be the funnel through which all information flows. And it’s not very good.”
The concern turned personal fast.
Brown said she thought, “My kids are going to be really dumb if we don’t figure out how to fix this.”
Legal pressure makes AI model accuracy a boardroom issue
Brown’s warning lands as companies push AI systems into hiring, lending, insurance, and other regulated decisions. Weak answers in those areas go far beyond user frustration. They trigger lawsuits, compliance failures, and reputational damage.
New York City’s Local Law 144 already requires employers using automated hiring tools to run bias audits, publish summaries, and notify job candidates. Reuters reported in October 2025 that companies using AI in hiring now face pressure to prove job relevance, monitor outcomes, and defend their decision models.
Brown said enterprises may drive the market toward stronger AI model accuracy because liability hits them directly when systems get things wrong.
Businesses using AI for credit, lending, insurance, and hiring, she said, “they’re going to want you to optimize for getting it right.”
She also dismissed weak compliance work in blunt terms, calling the current landscape “a joke.” Real evaluation needs domain experts, she said, not box-checking exercises.
“Smart generalists aren’t going to cut it.”
The search problem raises the stakes further
The accuracy problem also threatens publishers and readers. AI answers increasingly sit between users and original reporting.
Pew Research Center found in a March 2025 analysis that Google users clicked a traditional search result in only 8% of visits when they saw an AI summary. Without an AI summary, users clicked traditional results nearly twice as often, at 15%.
Fewer clicks mean less context. Less context means users rely more heavily on AI model accuracy — whether that accuracy deserves the trust or not.
Brown said she found troubling patterns inside current systems. She cited one major AI tool pulling from Chinese Communist Party websites “for stories that have nothing to do with China.” She also pointed to missing perspectives and weak handling of opposing arguments.
“There’s a long way to go,” Brown said. “But I also think that there are some very easy fixes that would vastly improve the outcomes.”
Brown sees Facebook’s old mistakes repeating
Brown spent years watching Facebook reward engagement over accuracy. She said those years taught her exactly what happens when platforms optimize for the wrong thing.
“We failed at a lot of the things we tried,” she said.
Social media already showed the cost of attention-first systems. Platforms chased clicks. Society paid with fractured trust and louder information wars. Brown now sees the same fork in the road ahead for AI companies.
“Right now it could go either way,” she said.
That choice touches classrooms, newsrooms, job searches, medical questions, and financial decisions. It also decides whether everyday users treat AI answers as reliable guidance or expensive noise.
Brown summed up the gap between tech industry promises and daily user experience with characteristic bluntness. Tech leaders speak about systems that will transform civilization, eliminate jobs, and cure disease, she said. Many users still get something far weaker.
“But then to a normal person who’s just using a chatbot to ask basic questions, they’re still getting a lot of slop and wrong answers,” Brown said.
The next leap in AI may not come from faster chips or sharper coding tests. It may come from whether companies decide to make AI model accuracy matter as much as speed, cost, and scale.
Should AI companies face tougher standards when their systems answer questions about news, politics, health, hiring, and finance? Please share your views about the problems with the AI model accuracy in the comments below.