As artificial intelligence becomes more integrated into how we access information, questions about who truly controls what AI tells us are more important than ever. Campbell Brown, a former news chief at Meta, has been vocal about the risks and responsibilities involved in guiding AI’s role in shaping public knowledge. She believes that currently, the industry is not doing enough to ensure AI provides accurate, nuanced, and trustworthy information. Her new venture, Forum AI, aims to address these challenges by creating standards for evaluating AI models on complex, high-stakes topics.

Why Accuracy in AI Matters

Brown’s background as a journalist and her experience at Meta have given her a clear perspective on the importance of truth. She watched social media platforms struggle with misinformation and now sees AI as the next frontier where misinformation can spread just as easily, if not more so. Her concern is that AI models often pull from biased or unreliable sources, leading to distorted or incomplete answers. This can have serious consequences when it comes to geopolitics, mental health, or financial decisions.

With Forum AI, Brown and her team are working to develop benchmarks that assess how well AI models perform on these sensitive topics. They bring in top experts—such as economists, former government officials, and cybersecurity leaders—to help set standards. The goal is to get AI systems to reach about 90% consensus with human experts, ensuring a higher level of reliability. This approach is meant to create AI that can be trusted to handle complex issues more responsibly.

The Challenges of Building Trustworthy AI

Brown recalls how she became alarmed after the release of ChatGPT, realizing it would be a primary channel for information. She observed that many AI models, including some from leading companies, still produce biased or inaccurate content. For example, some models draw from questionable sources or display political biases. These issues highlight the difficulty of creating AI that is both nuanced and truthful, especially when models often miss important context or perspectives.

She points out that fixing these problems requires more than just better algorithms. It demands domain expertise and careful evaluation of edge cases—scenarios that are not obvious but can cause serious misunderstandings or harm. Brown criticizes the current market for relying on checkbox audits and standardized benchmarks that she considers inadequate. She believes that genuine, nuanced evaluation takes time and expert work, which many companies neglect.

Her experience at Facebook showed her the dangers of prioritizing engagement over accuracy, leading to misinformation and less informed users. She hopes AI can break this cycle by focusing on truth rather than just clicks or engagement metrics. However, she admits that the industry is still figuring out how to make that shift, and skepticism among the public remains high. Trust in AI is at a low point, and many see it as unreliable or even dangerous when it comes to critical information.

Brown emphasizes that enterprise users—businesses involved in lending, insurance, hiring, and finance—have a strong incentive to demand more accurate AI. Companies worried about liability will want AI systems that get things right. Forum AI’s business model relies on this demand for responsible AI, but turning compliance into consistent revenue is still a challenge. She criticizes current legal and regulatory efforts for being superficial and missing the complexity of real-world scenarios.

She highlights the importance of expert evaluation, especially when assessing edge cases that can cause serious trouble if overlooked. For example, recent laws requiring AI audits for hiring practices have uncovered many violations that went unnoticed. Real evaluation requires deep domain knowledge and time—traits that few generalists can provide. Brown believes that only specialized experts can truly ensure AI models are safe and accurate for high-stakes applications.

With her company having raised $3 million recently, Brown is well-positioned to influence the future of AI transparency and accountability. She notes the disconnect between how tech leaders talk about AI’s potential and the reality faced by everyday users. While companies tout AI as a game-changer, most users still encounter unreliable answers. This gap fuels public skepticism, which she believes is justified given the current state of the technology.

Ultimately, Brown sees a path forward where AI can be guided by expert standards and responsible evaluation. She hopes that enterprises will prioritize truth and accountability, leading to a future where AI becomes a trustworthy source of information rather than a source of confusion. Achieving this will require industry-wide effort, transparent standards, and a commitment to accuracy—something she is actively working to promote through her new venture.

Inspired by

• https://techcrunch.com/2026/05/13/who-decides-what-ai-tells-you-campbell-brown-once-metas-news-chief-has-thoughts/

Sources

• byforum.com
• medium.com