California Bans AI-Only Firings and Requires Human Review
California will no longer allow employers to fire or discipline workers using AI alone. Governor Gavin Newsom signed the legislation on September 30, 2026, in a package his office framed as an extension of the state's AI regulatory framework. Engadget
What the law requires
Under the new law, an AI system cannot be the sole basis for a firing or disciplinary action. Human review is now mandatory. The rule applies to automated decision systems, meaning software used to hire, evaluate, schedule or manage workers, and it draws a clear line against fully automated firings. Engadget
Companies must also provide transparency when AI plays a causal role in larger workforce actions. If a mass layoff, relocation or termination is caused by an AI system, the employer must disclose that. A separate provision bars surveillance tools in workplace bathrooms. The protections are spread across multiple bills rather than a single statute. Engadget
How the state announced it
The Governor's Office published its announcement on September 30, 2026, under the title "California's nation-leading AI framework just got stronger — Governor Newsom signs more first-in-the-nation worker protections and more." Governor's Office The announcement positions the measures as worker-specific additions to existing state AI policy.
The case that came first
The signing follows litigation earlier this year over alleged use of AI in layoff selection. Former Meta employees alleged in a lawsuit filed in July that the company used AI to target workers with medical conditions for layoffs. Reuters The complaint was described as the first against a major U.S. company challenging the alleged use of AI in conducting layoffs. U.S. News
A U.S. judge declined on July 17 to block Meta from laying off 26 employees while they pursued that AI discrimination claim. Reuters The ruling addressed interim relief. It left the underlying allegations to be litigated separately.
What it means in practice
The broader context here is implementation, not principle. In my view, California is codifying a human-in-the-loop requirement, where a person checks the machine output before action, that many enterprise HR and workforce analytics vendors already describe as best practice. The hard questions will be familiar to anyone who runs machine learning in production. What counts as sole reliance when a score or ranking feeds a manager dashboard. What records prove real human review. How to trace cause when a planning model shapes a restructuring months before notices go out.
Worth flagging for technology leaders is the compliance surface. Performance management platforms, hiring systems, productivity monitoring tools and internal AI assistants now sit inside employment liability. Teams will need versioned models, reproducible decision trails and clear ownership between HR, legal and engineering. That work is unglamorous. It also makes systems more debuggable and defensible over time, which benefits employers as much as employees. I watched my own children enter the workforce as automated scheduling and screening became normal, and their expectation was simple. A person should be able to explain a decision that affects their livelihood. California has now written a version of that expectation into law.
