Ex-Meta News Chief Warns AI Risks Misinforming Users, Pushes for ‘Truth-Driven’ Systems

Who determines what artificial intelligence tells users? According to former Meta news executive Campbell Brown, the answer should increasingly lie with trusted experts rather than algorithms trained solely for speed and scale.

Brown, who once led news partnerships at Meta, is now focused on improving the reliability of AI-generated information through her startup, Forum AI. The New York-based company, founded about 17 months ago, aims to build systems that evaluate AI responses using expert-backed benchmarks.

The model, Brown explained, involves recruiting leading specialists to design standards for assessing AI accuracy and then training “AI judges” to review models at scale.

In Forum AI’s geopolitics work, the company has enlisted respected figures including historian Niall Ferguson, journalist Fareed Zakaria, former U.S. Secretary of State Tony Blinken, former House Speaker Kevin McCarthy, and cybersecurity expert Anne Neuberger.

Brown said the company’s objective is to achieve roughly 90 per cent alignment between AI evaluations and expert consensus.

Brown traced the inspiration for Forum AI to her time at Meta when OpenAI publicly released ChatGPT. She recalled quickly realising that AI chatbots could become the primary gateway through which people access information, but worried that the technology was not sufficiently reliable.

Concerned about the impact on future generations, Brown said she feared AI systems could negatively shape how young people learn if accuracy and quality were not prioritised.

She argued that many major AI firms remain heavily focused on coding and mathematical performance while paying insufficient attention to information quality, news accuracy and contextual understanding.

Although those challenges are harder to solve, Brown believes they are too important to ignore. Forum AI’s evaluations of major language models, she said, revealed troubling patterns, including sourcing concerns, political bias and missing context in responses.

Brown noted that subtle failures such as omitting key perspectives or misrepresenting arguments remain widespread, though she believes many of these issues are fixable.

Drawing from her years at Meta, Brown reflected on the consequences of platforms optimising for engagement rather than truth. She acknowledged that several initiatives she helped build, including Meta’s fact-checking programme, ultimately fell short.

Still, Brown hopes AI can avoid repeating the mistakes of social media by prioritising truthful and honest information over attention-driven outcomes.

While acknowledging that such a goal may sound idealistic, she believes businesses could become an important force for accountability, especially in sectors such as lending, hiring, insurance and credit decisions where accuracy directly affects legal and financial risk.

Brown also criticised the current state of AI compliance and auditing, describing many regulatory processes as inadequate. She argued that meaningful evaluations require deep domain expertise and careful testing of edge cases that generic assessments often overlook.

Despite bold promises from major technology firms about AI transforming industries and solving major global challenges, Brown said many ordinary users still encounter inaccurate or low-quality chatbot responses. As public trust in AI remains weak, she believes consumer scepticism is often justified, highlighting what she sees as a growing disconnect between Silicon Valley’s optimism and the everyday experience of users.