California moved on AI governance again. On April 1, 2026, Governor Gavin Newsom signed an executive order establishing guardrails for state employees’ use of artificial intelligence while encouraging accelerated adoption where appropriate, according to CalMatters.
The order has two distinct components, and separating them matters for compliance purposes.
Component 1: State agency AI governance
The operational core of the order directs California state agencies to develop recommendations for contract standards that address specific high-risk AI capabilities. Those capabilities, as confirmed by CalMatters, include AI’s potential to generate child sexual abuse material, violate civil liberties, and infringe protections against discrimination, detention, and surveillance.
This is not a regulation binding on private companies. It’s an executive directive for California’s own agencies. What it creates, practically, is a set of procurement standards that will govern which AI products state agencies can contract for. Companies selling AI services to California state government need to track what those standards look like when agencies publish their recommendations.
Component 2: Federal pushback
The more consequential piece, and the less confirmed one, is the order’s apparent response to a federal designation. According to CalMatters reporting, the order addresses a federal designation of Anthropic as a supply chain risk, directing California to conduct independent reviews of such federal determinations rather than deferring to them automatically.
The federal designation of a US AI company as a supply chain risk is an unusual claim. It has not been independently confirmed in this report beyond CalMatters’ coverage. Readers tracking this development should verify the underlying federal action through additional sources before drawing compliance conclusions from it. The direction of the political signal, however, is clear: California is asserting the right to make independent determinations on federal AI policy decisions that affect its own contractors and agencies.
Context: California’s recurring role
This executive order is the latest in a pattern. California has been the most active US state on AI governance, through legislation, veto, and now executive action. Governor Newsom vetoed SB 1047 in 2024. He has signed and vetoed a series of AI bills since. This EO operates differently: it doesn’t require legislative approval and takes effect immediately for state operations.
The White House Framework released March 20, 2026, explicitly calls for federal preemption of state AI laws imposing “undue burdens” on AI development. California just signed an order asserting independent authority in the same policy space. These aren’t compatible directions.
What to watch
State agencies must now develop the contract standard recommendations called for by the order, that process will produce the specific requirements. Watch for agency publications in the weeks following the order. The federal Anthropic supply chain designation, if confirmed, would represent a significant escalation of federal AI security policy and warrants independent monitoring.
TJS synthesis
The California EO is two stories in one. The state agency procurement standards are concrete and will produce compliance obligations for AI vendors selling to California government. The federal pushback dimension is significant if the underlying federal action is confirmed, but that confirmation matters before any organization acts on it. Don’t conflate the two. The procurement story is actionable now. The federal-state conflict story is worth watching, not yet worth building compliance posture around.