California's AI Layoff Bill Skips You. The One Beside It Doesn't.

California's Senate Appropriations Committee cleared a batch of AI employment bills on August 13, and two of them were amended the same day. The coverage that followed focused almost entirely on the one about layoffs.

For a small business, that is the wrong one to read.

Before going further: these are bills, not law. They cleared committee and were amended in mid-August. They still need floor votes and the Governor's signature, and bills die at both stages every year. Nothing here requires action this week. It is worth knowing now because one of them would reach small employers in a way the reporting has not made obvious.

The one everybody covered

SB 951 amends California's WARN Act to deal with technological displacement. When a mass layoff, relocation, or termination is caused in whole or in substantial part by an AI system or other automated technology, the employer has to disclose specifics: the number, classification, and work location of the affected roles, which job functions are being automated, and the category of AI system involved, including the entity that developed, sold, or leased it.

It also creates a new notice to the Employment Development Department for what the bill calls a technological cessation in hiring, meaning a permanent end to recruiting for a position because AI or automation now covers it. Violations carry civil penalties up to $500 per day. The bill was amended in the Assembly on August 13.

That naming requirement is unusual and worth noting. An employer would be disclosing to the state which vendor's product replaced the position.

Here is the part the coverage tends to skip. SB 951 operates inside Cal/WARN, and Cal/WARN applies to a covered establishment employing 75 or more people, currently or within the preceding twelve months. The notice period is 60 days.

If you employ fewer than 75 people, this bill is not about you. Most small businesses in the state are outside it entirely.

The one with no floor at all

AB 1883 was amended in the Senate on the same day, and it is built differently.

It prohibits employers from using workplace surveillance tools that can recognize, or make inferences or predictions about, a worker's emotional state, or that collect neural data. There are narrow exceptions for safety and for federal requirements tied to aircraft and national security work.

There is no employee threshold. None. The definition of employer reaches any entity controlling wages, benefits, or terms of employment, which takes in private businesses of every size alongside government entities, charter cities, the University of California, and school districts.

Penalties run up to $500 per violation. Enforcement runs three ways: the Labor Commissioner can investigate and cite, public prosecutors can act, and workers can bring their own civil actions for damages, injunctive relief, and attorney's fees.

That last clause is the one to sit with. A private right of action with fee recovery changes the economics of a small compliance mistake, because it does not depend on a state agency deciding your eleven-person company is worth its time.

Why an ordinary small business could get caught

The reasonable first reaction is that this has nothing to do with you, because you are not running emotion-detection software on your staff. Very few small businesses set out to.

The bill defines a workplace surveillance tool broadly: any system collecting worker data by means other than direct observation, including video, audio, time-tracking, geolocation, and biometric systems. That definition covers a great deal of ordinary software. Delivery route tracking. Call recording. The time clock app. Camera systems in a retail space.

The prohibition itself is narrower, and the distinction matters. What is banned is the emotional inference and the neural data, not the monitoring.

The exposure sits in the gap between those two facts. Vendors have spent three years adding sentiment analysis, engagement scoring, tone detection, and morale dashboards to products that were previously just recording hours or calls. Those features arrive in version updates. They get switched on by default. Nobody at a fifteen-person company reads the release notes for the scheduling tool.

A business could be running a prohibited feature without ever having decided to, inside a product it bought for an unrelated reason.

What is actually worth doing

Two things, and neither takes long.

Inventory what your monitoring tools claim they can do. Not what you use them for. What the vendor advertises. Pull up the feature list for anything that records, tracks, or scores employees, and look specifically for sentiment, mood, engagement, tone, stress, or wellbeing analysis. If a feature like that exists, find out whether it is on and who reads the output.

Ask vendors the question in writing. "Does this product infer or predict employee emotional state, and can that be disabled?" A written answer from the vendor is worth having regardless of what California does, because it is the kind of question that tends to get asked again later, by a different state or by your own insurer.

If you have no California employees and no plans for any, this is a watch item rather than a task. We would still do the inventory, for a practical reason: California employment rules have a long history of showing up elsewhere, and the inventory takes an afternoon whether you do it now or after a law names it.

The pattern worth noticing

The bill with the alarming headline has a threshold that excludes most small employers. The bill nobody wrote about has no threshold and lets employees sue.

That is not unusual. Size thresholds are where the real scope of an employment law lives, and they are almost never in the headline. When something new arrives, the first question worth asking is not what it prohibits. It is who it applies to, and whether the answer is everyone.

If you want a straight read on which of these actually touch your business, we are happy to look at it with you.