Campbell Brown, renowned for her career chasing accurate information as a TV journalist and Meta’s first dedicated news chief, is now confronting the profound impact of artificial intelligence on how people consume news. Witnessing the potential for history to repeat itself in the digital information landscape, Brown has launched Forum AI, a company dedicated to ensuring the integrity of AI-generated information.

Forum AI focuses on evaluating the performance of foundational AI models across what Brown terms “high-stakes topics,” including geopolitics, mental health, finance, and hiring. These are areas characterized by nuance, complexity, and a lack of simple yes-or-no answers. The company’s methodology involves enlisting world-leading experts, such as Niall Ferguson, Fareed Zakaria, former Secretary of State Tony Blinken, former House Speaker Kevin McCarthy, and Anne Neuberger, to architect rigorous benchmarks. Forum AI then trains AI judges to evaluate models at scale, aiming for approximately 90% consensus with human experts—a threshold it claims to have achieved. Brown’s motivation for founding Forum AI 17 months ago stemmed from the public release of ChatGPT, which she quickly realized would become a primary funnel for information, yet its initial accuracy was concerning, prompting worries about its implications for future generations. She expressed frustration that while foundation model companies prioritize coding and mathematics, the critical domain of news and information accuracy often falls by the wayside.

Initial evaluations by Forum AI have revealed significant shortcomings in leading models. Findings include instances of models like Gemini sourcing information from Chinese Communist Party websites for unrelated topics and a pervasive left-leaning political bias across nearly all models. Beyond overt inaccuracies, subtler failures such as missing context, omitted perspectives, and unacknowledged straw-man arguments are also prevalent. Brown acknowledges that there is a considerable journey ahead for AI accuracy, but also believes many straightforward fixes could vastly improve outcomes. Drawing on her experience at Meta, where optimizing for engagement often had detrimental societal effects and left users less informed, Brown hopes AI can break this cycle. She believes that while optimizing AI for truth might seem idealistic, enterprise demand could be a powerful catalyst. Businesses utilizing AI for critical decisions in credit, lending, insurance, and hiring are highly sensitive to liability and will inherently prioritize correctness.

Forum AI’s business model hinges on this enterprise need for accuracy, though converting compliance interest into consistent revenue remains a challenge in a market often satisfied with superficial audits. Brown critically assesses the current compliance landscape as inadequate, citing how a New York City law requiring AI audits for hiring bias found over half of violations undetected. She asserts that genuine evaluation demands deep domain expertise to navigate not only known scenarios but also complex edge cases that generalists often overlook. Despite the significant investment required, Forum AI, which raised $3 million last fall led by Lerer Hippeau, is uniquely positioned to address the stark contrast between the grand visions espoused by big tech leaders and the everyday reality for users who still frequently encounter erroneous information from chatbots. Brown maintains that the low levels of trust in AI are often justified, highlighting a significant disconnect between the industry’s internal discussions and consumers’ experiences.