AI doesn't need a regulator. It needs a referee.

The Washington Post published an opinion piece on May 22, 2026, titled "AI doesn't need a regulator. It needs a referee," penned by U.S. Representative Sam Liccardo. The article critiques the Trump administration's fluctuating approach to artificial intelligence governance, specifically noting the recent delay of a significant executive order intended to increase federal oversight of advanced AI models. This delay, following earlier proposals to penalize states with their own AI safety regulations and a subsequent suggestion for a federal vetting scheme, has created an environment of uncertainty in the AI policy landscape. Liccardo proposes an alternative framework, advocating for an "AI referee" system to incentivize the development of safe AI products. This system would leverage the inherent competitive drive of hyper-scalers and their desire to avoid liability. The core idea is to empower the Commerce Department's Center for AI Standards and Innovation (CAISI) to establish industry best practices for AI safety across various risk categories. Under this proposed model, AI companies whose models achieve or surpass CAISI's defined best practices would be granted federal preemption. This "safe harbor" would shield them from liability under potentially conflicting state-level AI safety laws for specific risk categories, such as cybersecurity, human safety, and privacy. The author argues that this approach would foster a "race to the top" among AI developers, as companies would strive for safer designs to gain a decisive regulatory advantage over their competitors. The article emphasizes that this referee system is better suited for a rapidly evolving technology like AI than rigid, fixed regulatory standards. While acknowledging that no law can entirely eliminate AI risk, Liccardo suggests this framework would encourage continuous improvement in AI safety. Other regulatory bodies, such as the SEC or FDA, would continue to oversee companies using AI within their respective sectors, focusing on outcomes rather than the tools themselves. This nuanced approach aims to balance innovation with necessary safeguards in the burgeoning AI industry.