California Governor Signs Executive Order Requiring AI Safety Certifications for State Vendors
Executive Order N-5-26 requires AI vendors seeking California state contracts to certify safeguards against illegal content, bias, and civil rights violations within 120 days.
Overview
California Governor Gavin Newsom signed Executive Order N-5-26 on March 30, establishing new trust and safety requirements for artificial intelligence companies seeking to do business with the state. The order directs state agencies to develop procurement certifications within 120 days that would require AI vendors to demonstrate safeguards against illegal content, harmful bias, and civil rights violations. It also instructs the state’s Chief Information Security Officer to independently review federal supply-chain risk designations rather than automatically deferring to federal determinations.
The executive order arrives amid a growing tension between state and federal approaches to AI regulation, with the Trump administration favoring a nationwide “light-touch” framework that does not address bias, discrimination, or civil rights concerns.
Procurement Certifications
The central mechanism of the order is a new vendor certification regime. Within 120 days, the California Government Operations Agency, in coordination with the Department of General Services and the California Department of Technology, must develop procurement certifications requiring entities that contract with the state to attest to policies addressing the exploitation or distribution of illegal content, harmful bias in AI models, and violations of civil rights and liberties, including surveillance, restrictions on free speech, and unlawful discrimination.
The same 120-day window applies to recommended reforms to contractor responsibility provisions. These reforms would prevent the state from contracting with entities judicially determined to have unlawfully undermined privacy or civil liberties, a provision that extends beyond AI-specific vendors to general state procurement.
Independent Federal Risk Review
One of the more notable provisions directs California to conduct its own assessment when the federal government designates a business as a supply-chain risk. This directive was prompted by the Department of Defense’s recent classification of San Francisco-based AI company Anthropic as a supply-chain risk after the company refused to remove contractual safeguards preventing military use of its systems for domestic mass surveillance and fully autonomous weaponry. A federal judge subsequently issued a temporary injunction blocking the designation.
Under the new order, the state’s Chief Information Security Officer may permit California agencies to work with vendors the federal government has flagged if the state determines the designation was improper.
AI Watermarking and Government Adoption
The order also addresses the growing challenge of AI-generated disinformation. Within 120 days, the California Department of Technology and Government Operations Agency must issue guidance on best practices for watermarking AI-generated or substantially manipulated images and video used by state agencies.
Alongside the safety provisions, the order directs agencies to expand their use of generative AI tools for public services. State employees are to receive access to vetted AI tools with privacy and cybersecurity safeguards, and agencies must pilot AI applications for government services, including a new tool to help Californians navigate state programs and benefits by life event.
State Versus Federal Approach
The executive order represents a direct challenge to the Trump administration’s preference for preempting state-level AI regulation. The White House unveiled a national AI framework earlier in March that favors minimal regulatory intervention. Newsom framed his approach as a necessary counterweight, stating that California would ensure AI companies protect people’s rights while the federal government pursues a more permissive stance.
Over 20 California state departments are already developing or using Poppy, a generative AI assistant, and multiple agencies are testing AI applications across various functions. The order seeks to accelerate this adoption while establishing guardrails that the federal framework currently lacks.