California SB 947 and AB 1883: What Changes When Newsom Decides by September 30

As of September 28, 2026, California SB 947 (the “No Robo Bosses Act”) and AB 1883 (the AI emotion-and-neural-data surveillance ban) are both sitting on Governor Newsom’s desk, unsigned and unvetoed, with a constitutional deadline of September 30, 2026. Both passed the Legislature with bipartisan margins and would reshape how California employers use automated systems to discipline, fire and monitor staff.

This post lays out exactly what each bill requires, what changes under every possible outcome, and what a training, HR or compliance team should build now regardless of which way Newsom goes. Nothing here is legal advice; check with employment counsel before changing a policy.

What Are SB 947 and AB 1883, in Plain English?

SB 947 stops employers from firing or disciplining a worker based solely on an automated decision system, requiring a human reviewer to corroborate the outcome first. AB 1883 bans AI-based workplace surveillance tools that infer an employee’s emotions or collect neural data. Both are pending gubernatorial action with a September 30, 2026 deadline.

Senator Jerry McNerney (D-Pleasanton, SD-5) authored SB 947 after a nearly identical bill, SB 7, died on Newsom’s desk in October 2025. Assemblymember Rebecca Bauer-Kahan’s AB 1883 moves independently but covers adjacent ground: where SB 947 governs the decision (should a person get fired), AB 1883 governs the input (can an AI tool read that person’s face or brainwaves to help decide).

Neither bill is theoretical for training platforms. SB 947 reaches any system that scores, ranks or recommends discipline based on training completion, assessment performance or behavioral analytics. AB 1883 reaches webcam-based proctoring and engagement-scoring features that infer mood or attention during e-learning.

What Is the Current Status of SB 947 and AB 1883 Right Now?

Both bills remain unsigned and unvetoed as of this writing. SB 947 was enrolled and presented to the Governor on September 9, 2026; AB 1883 followed on September 10, 2026. Neither appears in Newsom’s legislative updates of September 20 or September 27, which together logged hundreds of signings and vetoes without touching either bill.

That silence is not unusual this late in the cycle. Governors routinely hold the most contested bills for the final days of the signing period, and both of these carry organized opposition from employer groups and organized support from labor, which tends to push a decision to the wire. Watch gov.ca.gov’s newsroom for the next legislative update, which should land before or on September 30.

What Would SB 947’s Human-Review Rule Require of an LMS or HR Platform?

SB 947 requires that when an automated decision system is a primary basis for discipline or termination, a human reviewer with authority to reverse the outcome must corroborate it, and the affected worker must get written notice and access to the data used against them. A system that only assists a human who already has the full picture is not what the bill targets.

The bill also bars automated systems from predicting a worker’s behavior, beliefs or personality, or from identifying employees who exercise legally protected rights such as union organizing. For any platform whose analytics dashboard flags “at-risk” or “low-engagement” learners in ways that feed into performance write-ups, that scoring logic becomes a compliance surface, not just a feature.

Enforcement runs through the Labor Commissioner and public prosecutors, with civil penalties of $500 per violation and a private right of action, meaning affected workers can sue directly. Multiply that across a workforce and a poorly documented rollout of AI-assisted performance management gets expensive fast.

What Counts as an “Automated Decision System” Under SB 947?

SB 947 defines an automated decision system broadly, as any computational process that materially informs a discipline or termination decision, regardless of whether it uses machine learning. This mirrors the definition California’s Civil Rights Council already adopted for its automated-decision-system regulations under the Fair Employment and Housing Act.

That means the definition doesn’t hinge on marketing language. A learning platform’s built-in “performance risk” flag, a third-party analytics add-on, and a home-grown spreadsheet formula that scores attendance can all qualify if the output shapes who gets a warning or a pink slip.

Which E-Learning and Proctoring Features Would Become Risky Under AB 1883?

AB 1883 bans employers from deploying AI tools that recognize, infer or predict an employee’s emotional state, or that collect neural data from the nervous system, with penalties of $500 per violation. It defines “workplace surveillance tool” broadly enough to cover video, audio, geolocation and biometric monitoring wherever AI does the emotional inference.

For a training organization, the exposure sits squarely in remote proctoring and engagement analytics. A proctoring tool that flags “suspicious” behavior from posture or gaze tracking is probably fine; one that scores a learner’s frustration, boredom or confidence level from facial expression crosses into banned territory the moment AI is doing the inferring.

The bill does not ban surveillance outright. Time tracking, geolocation for field staff and safety-related video monitoring stay legal as long as no AI layer is reading emotional state or neural signals into the output. The distinguishing question is not “are we watching people” but “is a model claiming to know how they feel.”

What Happens Under Each of the Four Possible Outcomes by September 30?

Four combinations are possible before the deadline: both bills signed, only SB 947 signed, only AB 1883 signed, or neither signed. Each produces a different compliance timeline, and under California’s constitutional rule a bill Newsom simply ignores becomes law anyway, without his signature, once the deadline passes.

How Does SB 947 Compare With Colorado’s AI Human-Review Rule?

Colorado’s amended AI Act takes a reactive approach: AI can make or heavily influence an employment decision, but an independent, trained human reviewer must reconsider any adverse outcome within 45 days if the worker requests it. SB 947 takes a preventive approach: it bars sole reliance on the automated system before the decision is finalized, not after.

Colorado’s rulebook, effective January 1, 2027, spells out exactly what “meaningful” review means. The reviewer cannot be the original decision-maker or their direct supervisor, must be trained on the factors the system weighed, must have real authority to reverse the outcome, and cannot use AI to help conduct the review itself. Documentation of the reviewer’s identity, training and reasoning has to be retained.

A vendor or employer operating in both states will end up needing the stricter of the two workflows almost everywhere: pre-decision human sign-off to satisfy California, plus a documented post-decision reconsideration path to satisfy Colorado. Building one process that covers both is cheaper than building two.

What Happens if Newsom Takes No Action at All by the Deadline?

Unlike the federal “pocket veto,” California has no mechanism for a bill to quietly die from inaction. Under Article IV, Section 10 of the state constitution, any bill Newsom neither signs nor vetoes by September 30 becomes law automatically, gets a chapter number, and takes effect on its stated operative date exactly as if he had signed it.

That matters here because both bills carry organized opposition, and a governor who wants to avoid a signing ceremony photo with labor unions, or a veto message that angers them, has a third option: say nothing and let the clock run out. It is a real, historically used outcome, not a procedural footnote.

Why Did Newsom Veto a Similar Bill Last Year?

In October 2025, Newsom vetoed SB 7, the prior year’s “No Robo Bosses Act,” calling it unfocused and warning it could sweep in routine, low-risk automated tools alongside genuinely consequential ones. McNerney’s team narrowed SB 947’s scope in response, tightening the definition of automated decision system and limiting the human-review trigger to discipline and termination rather than a broader set of employment actions.

That history is the main reason this year’s outcome is genuinely uncertain rather than a formality. A narrower bill addressing a governor’s stated objections often fares better on a second attempt, but Newsom, in the final stretch of his term, has also shown willingness to veto business-costly bills even after industry-requested amendments.

What Should L&D and HR Teams Prepare Now, Regardless of the Outcome?

Regardless of which way Newsom decides, three moves reduce risk under either bill and under the existing FEHA automated-decision-system rules already in force. Start by inventorying every system that scores, flags or ranks employees, then decide who has final sign-off authority, then document that reviewer’s training.

Build a written human-review workflow now. Map every point where a training or performance system’s output feeds a disciplinary decision, name the human reviewer, and record what evidence they examined beyond the algorithm’s summary. This is the same documentation Colorado now requires and the same evidence a California plaintiff’s attorney would ask for first.

Separate “monitoring” from “inferring.” Audit proctoring and engagement-analytics tools for any feature that claims to detect emotion, attention level, confidence or stress from video, audio or biometric signals, and get a straight answer from the vendor about what model powers it. A documented LMS governance framework that assigns clear ownership over these questions makes this audit repeatable instead of a one-off scramble.

Fold this into existing compliance infrastructure rather than building a parallel one. Most organizations already run training compliance management software to track completions and certifications; the same system of record can hold reviewer sign-offs and audit trails for AI-influenced employment decisions. Pair that with AI agent governance training for the managers and HR business partners who will actually be doing the reviewing, since the rule is only as good as the person applying it.

Does This Apply Outside California?

No, SB 947 and AB 1883 would only bind employers with California employees, but the compliance work travels. Any national platform that builds a human-review workflow to satisfy California will likely reuse it for Colorado, and increasingly for New York City’s Local Law 144 and Illinois’s biometric and AI-hiring statutes, since the underlying documentation is nearly identical across all four.

Read more: What Connecticut’s Public Act 26-15 Requires Employers to Disclose About AI-Caused Layoffs 

Vendors selling AI features inside an LMS to multi-state employers should treat California’s rules as the practical floor, not a regional edge case, given how much of the U.S. workforce touches a California-based employer at some point.

How Does This Interact With California’s AI Transparency Act?

It does not overlap directly. California’s AI Transparency Act (SB 942) governs disclosure and provenance labeling for AI-generated content shown to consumers, such as watermarking synthetic media, and has nothing to do with employment decisions or workplace surveillance. SB 947 and AB 1883 sit in employment law, not consumer-protection disclosure law, so a compliance program for one does not automatically satisfy the other.

An organization already building LMS-based compliance training programs for SB 942 disclosure obligations will need a separate, employment-specific track for SB 947 and AB 1883, even though the same governance team likely owns both.

Conclusion

Both bills are still live as of this writing, and California’s no-pocket-veto rule means silence from Newsom by September 30 is functionally the same as a signature. Don’t wait for a press release to start the work.

Inventory the automated systems already touching discipline, termination and monitoring decisions today, name a human reviewer for each, and get your proctoring and engagement-analytics vendors on record about what their models actually infer. Whatever Newsom decides, that groundwork holds up under FEHA’s existing rules, Colorado’s, and whichever version of SB 947 and AB 1883 lands.

Check the bill status directly once the deadline passes, since a chapter number can appear without a signing ceremony, and revisit your training and review workflows against the confirmed effective dates rather than the version reported in the news this week.