In this episode of HELLO FUTURE, host Kevin Cirilli speaks with David Fuscus, CEO of Précis AI and creator of Willard AI, about the most dangerous flaw in artificial intelligence for the policy world: hallucinations and inaccuracies. In a town where a single wrong citation can kill a bill, Fuscus explains why general AI models remain unreliable — pointing to OpenAI’s own findings that newer models sometimes hallucinate at higher rates, and Stanford research showing chatbots get legal questions wrong more than half the time and tend to agree with false statements from users.

The conversation digs into how Willard addresses this by wiring directly into federal government databases rather than relying solely on internet training data, and why the ability to trace every answer back to its source is essential for lobbyists and Hill staff. Fuscus also addresses growing concerns about AI flooding the system with synthetic public comments — including a California case that generated more than 20,000 AI-written submissions against clean-air rules — and whether lawmakers can still tell real constituent voices from manufactured noise. 

The episode closes with practical guidance for everyday listeners on when to trust tools like ChatGPT and when to treat their answers with extreme caution.