Skip to content
Welcome, thoughtbot's Giant Robots listeners!Start here
AI: Jobs, Power & Money
21SEP

California bill sets 90-day AI layoff notice

3 min read
16:45UTC

California's SB 951, requiring 90 days' notice before AI-driven displacement of a quarter of a workforce, cleared the Senate 28-9 and advanced through the Assembly this week.

EconomicDeveloping
Key takeaway

California advanced a 90-day AI-layoff notice that would outpace the federal WARN floor.

California's SB 951, requiring 90 days' notice before AI-driven displacement of 25% or more of a workforce, cleared the state Senate 28-9 on Wednesday 20 May 2026 and advanced through the Assembly committee stage in the week ending 6 June⁠1. Companion bill SB 947, addressing further worker protections on AI deployment, passed the Senate 29-9 the day before, on Tuesday 19 May, and is also moving through the Assembly.

The 90-day figure is the operative detail. It exceeds the federal Worker Adjustment and Retraining Notification (WARN) Act, which sets a 60-day floor for mass-layoff notice at employers of 100 or more. Because large companies run multistate operations and cannot easily apply different notice periods by location, the strictest state rule tends to become the de facto national standard. If SB 951 reaches the governor's desk and is signed, its 90-day clock would in practice govern AI-driven cuts well beyond California.

Oracle's offshore terminations and remote-reclassification tactics showed how the 60-day requirement can be sidestepped, with US WARN filings covering under 4% of its workforce cuts. California is also moving against the grain of Colorado, which retreated to a notice-only framework in May. The state-by-state divergence, tightening in Sacramento and loosening in Denver, is now the operative pattern in the absence of any federal AI-workforce law.

Deep Analysis

In plain English

California's state Senate voted 28 to 9 in favour of a bill called SB 951 that would require companies to give workers 90 days' notice before using AI to eliminate 25% or more of the workforce. The bill advanced through Assembly committees in the week of 6 June and is continuing through the legislative process. The US federal law on mass layoffs, called the WARN Act, only requires 60 days' notice. California's bill sets a longer window and applies a different trigger: it counts the total company headcount rather than the number affected at a single location. That matters because some companies have split roles across multiple offices specifically to avoid triggering the existing law. Colorado tried to pass a stronger version of this kind of law in 2024, but it was challenged in court and weakened. California's bill will likely face a similar legal challenge if it becomes law.

Deep Analysis
Root Causes

The bill's 90-day trigger directly addresses Oracle's WARN Act avoidance: the remote-reclassification tactic that allowed Oracle to file Massachusetts WARN notices covering under 4% of its affected workforce directly motivated the site-independent trigger mechanism.

Leading the Future is spending over $100 million to defeat regulation-minded candidates in 2026 primaries, creating an incentive for legislators in safe districts to pass protective legislation before the November general election shifts the chamber composition.

The federal measurement gap provides a third pressure: with the BLS having skipped its GenAI workplace paper since April 2026, state governments face mounting urgency to create their own data infrastructure rather than wait for federal action.

What could happen next?
  • Precedent

    If SB 951 becomes law and survives a constitutional challenge, it creates the first enforceable AI-displacement notice standard in US employment law, potentially displacing the federal WARN Act as the effective floor for tech-sector restructuring.

  • Risk

    A federal constitutional challenge using Colorado's DOJ-backed precedent could stay SB 951 before enforcement begins, creating a two-year legal limbo that allows AI restructuring to continue without regulatory constraint.

First Reported In

Update #12 · Jobs report says fine, layoff report says no

Transparency Coalition· 8 Jun 2026
Read original →
Different Perspectives
Korea's four largest banks, as reported by News1
Korea's four largest banks, as reported by News1
KB Kookmin, Shinhan, Hana and Woori rehired 1,013 retirees last year, equal to 79.1% of their graduate intake, according to News1. Graduates face a smaller entry intake while returners take fixed-term credit-screening roles at lower pay, News1 reports.
High Court of Justice of Madrid
High Court of Justice of Madrid
The court's employment chamber ruled in judgment 864/2026, dated 8 October and made public in early November, that introducing AI is not on its own a ground for dismissal. Employers must prove AI's effect through the four causes the Workers' Statute already lists.
Singapore Ministry of Manpower
Singapore Ministry of Manpower
The Singapore Ministry of Manpower put third-quarter retrenchments at 4,800, concentrated in information and communications, financial services and manufacturing, with business reorganisation the main driver. Its release does not mention AI.
Challenger, Gray & Christmas
Challenger, Gray & Christmas
Challenger reported on 5 November that market and economic conditions overtook AI as the leading stated reason for 2026's US job cuts, at 130,531 against 121,401. AI was cited for 1,265 cuts in October, its lowest month of the year.
Office for Budget Responsibility
Office for Budget Responsibility
The OBR's 28 October outlook showed output per worker in information and communication growing 7.6% a year, yet kept AI's boost to UK productivity at 0.2 percentage points a year. It does not attribute the faster growth to AI.
Central Bank of Ireland
Central Bank of Ireland
Governor Gabriel Makhlouf said on 2 November that AI investment has supported Irish activity and flagged the debt financing behind it as a financial-stability risk. On jobs, he named skills bottlenecks and mismatches, not job losses.

Written with AI from Lowdown's published reporting. Not individually reviewed by an editor. Editorial standards.