While Congress debates whether AI needs federal oversight at all, California just started building the machinery to enforce it.
In an interview with Reuters published Tuesday, California Attorney General Rob Bonta revealed that his office is constructing a dedicated "AI accountability program" — a permanent enforcement unit specifically designed to investigate and prosecute AI-related harms. And its first active investigation is already underway: a probe into Elon Musk's xAI over the generation of non-consensual sexually explicit images by its chatbot, Grok.
This isn't a task force. It's not an advisory committee. It's a permanent prosecutorial unit. And it signals that the state where most major AI companies are headquartered has decided not to wait for Washington.
The xAI Investigation
Bonta's office previously issued a cease-and-desist letter to xAI after reports that Grok was generating sexualized images — including images that appeared to depict minors — without consent safeguards. The investigation has continued, with Bonta telling Reuters that his office is actively looking into whether xAI violated California's existing consumer protection and privacy laws.
The probe operates on California's current legal framework, not on any new AI-specific legislation. That's notable. Bonta isn't waiting for lawmakers to pass an AI bill — he's using existing authority over consumer fraud, unfair business practices, and privacy violations to go after an AI company now.
This matters because the US has no federal AI regulation. The EU has the AI Act. China has its own generative AI rules. South Korea just passed comprehensive AI legislation in January. The United States has a patchwork of state laws and executive orders that don't add up to a coherent framework. California is filling the gap — not with legislation, but with enforcement.
Why a Permanent Unit
The AI accountability program Bonta described goes beyond the xAI case. It's designed to be a standing capability within the Attorney General's office — a team with the technical expertise and legal authority to investigate AI companies on an ongoing basis.
Think of it as California's answer to a structural problem: AI harms are multiplying faster than investigations can be opened. Deepfake abuse, algorithmic discrimination, data privacy violations, deceptive AI-generated content — each requires technical literacy that most law enforcement offices don't have. A dedicated unit means investigators who understand transformer architectures reviewing complaints from people who don't.
California has precedent for this approach. Its Environmental Justice Bureau, its Privacy Protection Unit, its Antitrust Section — all are permanent offices that give the AG capacity to investigate specific categories of harm without starting from zero each time. The AI accountability program follows the same model.
The Federal Vacuum
Bonta's move is, implicitly, a statement about federal inaction. Congress has held dozens of hearings on AI. Multiple bills have been introduced. None have passed. The closest thing to federal AI policy is a series of executive orders that can be — and have been — reversed with each administration.
Meanwhile, the AI regulation battle is now entering the political advertising space. As The Hill reported this week, millions of dollars are flowing into midterm election ads on both sides of the AI regulation debate. Tech industry groups are running ads warning that regulation will kill innovation. Consumer groups are running ads about deepfakes and job displacement. The fight over AI governance in the US is becoming a campaign issue — which means it's becoming slower, more polarized, and less likely to produce coherent law.
In that vacuum, California's AG is doing what California has done before on environmental regulation, data privacy (CCPA), and auto emissions: acting alone and hoping the rest of the country follows.
What to Watch
- Whether xAI cooperates or fights. The company's public posture — calling media reports "lies" — suggests a combative approach. California's AG has subpoena power. That's a mismatch xAI may not be prepared for.
- Whether other state AGs follow. New York, Illinois, and Texas have all expressed interest in AI enforcement. A coalition of state-level AI units would create a de facto national framework without Congress lifting a finger.
- The midterm effect. If AI regulation becomes a campaign issue, it could either accelerate federal action or freeze it entirely. History suggests the latter.
Why This Matters
The most consequential AI governance in the United States may not come from Congress, the White House, or any federal agency. It may come from a state attorney general's office in Sacramento with a small team, existing consumer protection law, and the determination to use it. California just built permanent infrastructure to hold AI companies accountable. Whether that's enough depends on whether anyone else builds their own — or whether Washington finally wakes up.