California adopts flexible AI oversight model amid rising compliance costs

Earlier this week, California passed legislation that authorizes Independent Verification Organizations to test cutting-edge AI models prior to their release, seeking to mitigate associated hazards. The law arrives amid skepticism that conventional AI regulation is both impractical and financially burdensome, exemplified by a METR probe into an OpenAI-Hugging Face breach that required $400,000 in AI token expenses. The author argues that static, hastily drafted bills quickly become obsolete as AI systems grow in scale, often needing other AI tools for analysis.

Instead, a dynamic governance model—crafted by knowledgeable experts willing to adjust course—would better address the rapid pace of innovation. Without such flexibility, regulations risk being ineffective or counterproductive, a scenario familiar to Californians from past policy failures.