California has disclosed the use of six high-risk artificial intelligence systems across state agencies — a jump from zero such systems disclosed last year. The announcement, made public this week, marks a significant shift in the state's approach to AI governance and shines a light on where the technology is being deployed in ways that could affect people's lives.
From zero to six: what changed
Last year, California reported no high-risk AI systems in use. That number now stands at six. The disclosure is part of the state's broader effort to track and regulate AI tools that could pose risks to residents' rights, safety, or access to services. While the state hasn't named the specific systems, the increase alone signals that agencies are now identifying and categorizing AI tools they previously may not have flagged.
Why the jump? State officials point to updated reporting requirements and a growing awareness of what qualifies as high-risk. The systems in question likely involve decisions about benefits, law enforcement, or health care — areas where an AI's error could have serious consequences.
Transparency and the trust gap
Public trust in government AI has been shaky. A lack of clear information about how algorithms work — and who they might harm — fuels that skepticism. California's disclosure aims to change that by offering a baseline: here are the systems, and here's the risk level.
But transparency alone isn't enough. Advocacy groups have pushed for independent audits and a mechanism for residents to challenge AI-driven decisions. The state's move is a step, but it doesn't yet include those safeguards. That leaves questions about accountability unanswered.
What this means for policy
California has been a testing ground for AI regulation, with lawmakers considering dozens of bills aimed at governing the technology. The disclosure gives policymakers concrete data to work with. Instead of debating hypothetical risks, they can now point to six real systems — and ask whether they're being used properly.
The number also sets a benchmark. If California can track high-risk AI, other states likely can too. The disclosure model could become a template for federal legislation, though no national standard exists yet.
The state is required to update its AI inventory annually. That means by next year, the public will see whether the number grows — or whether agencies are removing systems from the high-risk category as they improve oversight. For now, the six systems remain unnamed, and the full risk assessments haven't been released. Those details are expected to come in a follow-up report later this year.
