California has just passed an AI safety bill, SB 1047. The core requirements are straightforward: companies developing large AI models must conduct safety tests and establish accountability mechanisms. Violators may face civil penalties.

Supporters of the bill argue it’s a responsible approach. "We don’t want to stop innovation; we want to ensure it’s safe," said State Senator Scott Wiener. Opposition comes from some startups and advocacy groups. Adam Billen, VP of Policy at Encode AI, stated bluntly: "Will bills like SB 53 stop us from surpassing China? No. Calling this a hurdle in the race is intellectually dishonest."

The debate centers on regulatory costs. The bill provides exemptions for small developers meeting specific criteria, but critics argue compliance burdens remain too heavy.

Observations suggest this dichotomy is overstated. Regulation and innovation are not a zero-sum game. Clear rules can reduce uncertainty and direct resources to truly valuable areas. California’s example shows the discussion can be more pragmatic: not whether to regulate, but how to design smarter rules.

Image: Exterior of the California State Capitol, labeled "SB 1047 Passed Here"

The reality is that AI is advancing too rapidly for complete laissez-faire or excessive restrictions to be viable. The key is finding a balance between safety and progress. California’s experiment is worth watching—it may offer a model for other regions. After all, true innovation doesn’t require sacrificing safety.