California SB 7 "No Robo Bosses" Act 2026: AI Employment Decision Compliance
California's algorithmic-management push isn't aimed at resume screeners. It's aimed at the software already running your warehouse floor, your delivery routes, and your call queue — systems that quietly decide who gets scheduled, flagged, coached, and cut. Most employers have never classified those tools as employment-decision systems, which is exactly the exposure.
The Scope Problem: Your Scheduling Software Is an Employment Decision System
When employers hear "AI employment law," they audit their applicant tracking system and stop. California's framing is much broader: an automated decision system is any computational process that materially facilitates a decision about a worker. That sweeps in shift-assignment algorithms, productivity-rate scoring in fulfillment centers, telematics-based driver safety scores, adherence and sentiment scoring in contact centers, and any tool that auto-generates a performance flag a manager then acts on.
None of those are sold as HR technology. They're sold as operations tooling, bought by operations leaders, and never routed through employment counsel. That procurement path is the single best predictor of which companies get caught flat-footed.
What's In Scope vs. What Usually Isn't
- •Resume screening, ranking, and knockout question scoring
- •Automated scheduling and shift-assignment algorithms
- •Productivity quota and rate-tracking systems
- •Driver telematics and safety scoring tied to discipline
- •Call-center adherence, QA, and sentiment scoring
- •Automated PIP triggers and termination recommendations
- •AI-assisted promotion and compensation banding
- •Payroll processing and benefits administration mechanics
- •Job-board posting distribution
- •Anonymous aggregate engagement surveys
- •Security access provisioning not tied to performance
- •Training content recommendation with no evaluative output
"Human In The Loop" Has to Mean Something
The no-sole-reliance principle is where most compliance programs are thinnest. Every employer claims a human reviews the output. Under scrutiny, that review is often a manager clicking approve on a queue of eighty system-generated flags, with no access to the underlying inputs and no realistic ability to disagree. That is rubber-stamping, and it does not satisfy a human-review requirement in any of the frameworks now in play — California's, Colorado's, or the EU's.
A defensible review means the reviewer sees the factors that drove the output, has authority to override without escalation, and leaves a record of what they considered. If your tooling can't show a manager why a worker was flagged, the manager cannot perform the review the law contemplates — and that's a vendor requirement, not a training problem.
How It Stacks With FEHA, Local Law 144, and the Colorado AI Act
Multi-state employers should stop treating these as separate projects. California's Civil Rights Council regulations under FEHA took effect October 1, 2025 and already impose anti-discrimination and recordkeeping duties on automated decision systems. New York City's Local Law 144 requires annual independent bias audits with published results for automated employment decision tools. Colorado's AI Act layers duty-of-care and impact-assessment obligations on consequential decisions. The overlapping core — inventory, bias testing, notice, human review, retention — is roughly eighty percent of the work in every jurisdiction. Build it once against the strictest requirement and map the deltas.
Employer Compliance Checklist
Immediate Actions
- ☐Inventory every system that scores, ranks, or flags workers — including ops tools
- ☐Identify which outputs feed discipline, scheduling, or termination
- ☐Ask each vendor in writing whether they provide per-decision explanations
- ☐Check whether managers can actually override, or only approve
Build This Quarter
- ☐Written notice to workers describing ADS use and the decisions it informs
- ☐A documented human-review step with override authority and a record trail
- ☐Bias testing on any tool influencing hiring, promotion, or termination
- ☐Four-year retention of ADS inputs, outputs, and review records
- ☐Route operations-software procurement through employment counsel
Frequently Asked Questions
What is the 'No Robo Bosses' Act?
The nickname for California's SB 7 effort to regulate automated decision systems in the workplace: notice to workers when an ADS is used on them, no sole reliance on it for consequential decisions like discipline and termination, and a route to human review.
Does this only apply to AI hiring tools?
No. Scope runs across hiring, promotion, compensation, scheduling, discipline, and termination. Warehouse quota systems, driver telematics scoring, and automated PIP triggers are in scope even though no one calls them HR software.
How does it interact with California's FEHA automated-decision regulations?
They overlap. The Civil Rights Council's FEHA regulations, effective October 1, 2025, address ADS as a discrimination source and extend recordkeeping. SB 7-style rules add notice, human review, and no-sole-reliance on top. One compliance build substantially serves both.
Are AI vendors liable, or only employers?
Under the FEHA regulations, agents acting on an employer's behalf — staffing firms, screening vendors — can be treated as employers for liability. But notice and human-review duties toward your workforce stay with you, so contracts must guarantee the explainability you need.
What records should we keep?
ADS inputs and outputs, selection criteria, bias-testing documentation, worker notices, and proof a human reviewed and could override each consequential decision — retained four years. An undocumented review is treated as no review.
We're headquartered outside California but employ Californians remotely. Does it reach us?
California employment protections generally follow the worker's place of work. Remote employees based in California pull their employer into California's framework for decisions affecting them, regardless of where the company or the software is headquartered.
Start With the Inventory, Not the Statute
The legislative details will keep moving. What won't change is that you cannot provide notice, human review, or bias testing for systems you haven't identified. Every employer that gets caught here gets caught the same way: an operations tool nobody classified as an employment-decision system quietly drove a termination.
The same blind spot applies to what your public career site does to applicants with disabilities. Scan your careers pages for accessibility barriers that screen people out before any algorithm does.