While the federal government continues to twiddle its thumbs on AI regulation, California decided to actually do something about it. Governor Newsom signed SB 53 into law, establishing new standards for frontier AI models that focus on safety and transparency. Because apparently someone needs to be the adult in the room.

While Washington sits idle on AI oversight, California stepped up with actual regulations because somebody had to take charge.

The law targets what they call "frontier models" – AI systems trained with computing power exceeding 10^26 FLOPs. That's a lot of computational muscle. But it doesn't apply to every garage startup tinkering with AI. Only large frontier developers with annual revenues over $500 million need to worry. So your neighborhood AI enthusiast can relax.

Here's what these big players have to do now. They must publish transparency frameworks detailing how they incorporate national and international AI standards. No more black box mystery approaches. The law also creates CalCompute, a consortium designed to foster AI research and innovation. Because California loves its consortiums.

Safety takes center stage too. A new reporting system for critical safety incidents gets housed under the Office of Emergency Services. Developers must report incidents presenting imminent risks within 24 hours to relevant public safety authorities. Whistleblower protections exist for those brave enough to speak up about problems. Civil penalties await companies that ignore the rules. Real consequences, imagine that.

The employment side gets interesting. Starting October 1, 2025, anti-discrimination laws extend to AI tools used in hiring decisions. Automated decision systems become potential agents of employers. That means AI can't discriminate based on race, gender, or other protected characteristics. Revolutionary concept, really. Employers must maintain record-keeping of AI decision-making data for a minimum of four years to ensure compliance. Regular bias checks remain critical to prevent discriminatory outcomes that could harm specific demographics.

California's approach exceeds even the EU's AI Act threshold for defining frontier models. The state isn't messing around with stringent requirements. The California Department of Technology gets tasked with recommending updates as technology advances. Because these definitions won't mature well in the fast-moving AI world.

This law serves as a blueprint for national policies, given the federal government's apparent inability to address AI regulation thoroughly. California often leads where Washington fears to tread. Whether other states follow suit remains to be seen. But at least someone's taking the wheel while AI development accelerates at breakneck speed.