Campbell Brown, former head of Meta's news division, has launched Forum AI to evaluate the performance of AI models on complex topics such as geopolitics and finance. Brown highlights that while chatbots are becoming key information gateways, their responses on high-risk topics remain inconsistent. She argues that the industry prioritizes programming over accuracy, leading to unreliable answers. Forum AI's approach involves domain experts designing evaluation benchmarks, with AI systems scoring model responses. Brown notes that early evaluations reveal issues like political bias and lack of context. She emphasizes the need for accurate AI responses, especially in sectors like credit and hiring, where errors pose liability risks. Brown also criticizes current AI compliance audits as superficial, advocating for expert involvement in evaluations.