California just made it illegal for an AI alone to fire you

SB 947, nicknamed the No Robo Bosses Act, prohibits California employers from relying exclusively on an automated decision system to fire or discipline a worker. Before acting on an AI-generated recommendation, a human has to independently review and corroborate it using actual supporting information, like managerial evaluations, peer reviews, or personnel files, and must reject any output found to be inaccurate, incomplete, or misleading.

• Human review required before any AI-driven firing or discipline decision takes effect.
• Written notice to affected workers, including contact info for the human reviewer.
• Data access rights, letting workers request a description of the data used about them.
• Penalty: $500 per violation, plus possible punitive damages and attorney's fees.
• Effective date: July 1, 2027.

Why it matters: this doesn't ban AI from employment decisions, it specifically blocks a company from letting an algorithm be the sole, unverified basis for ending someone's job. The gap between "AI assists the decision" and "AI is the decision" is exactly what this law targets.

Background: Newsom vetoed the earlier version, SB 7, in October 2025, explaining that while he shared concerns about employer use of AI, the bill imposed "unfocused notification requirements" that would have applied to virtually any business using even the most routine digital tools, things like basic scheduling or workflow management software, not just AI making real employment decisions.

Why it matters: that veto wasn't a rejection of the underlying concern, it was a rejection of overly broad drafting. Lawmakers took that feedback seriously enough to substantially narrow the bill's scope before bringing it back.

In real life it's the legislative equivalent of a first draft trying to ban "anything with a computer chip in it" and a second draft narrowing that down to "specifically, the thing that decides whether you keep your job."

The revised bill includes a narrower definition of what counts as an automated decision system, focuses specifically on termination and discipline decisions rather than broad digital tool usage generally, and was designed to interact more cleanly with California's existing Fair Employment and Housing Act protections rather than creating overlapping, confusing rules.

Why it matters: that narrower scope is likely exactly why this version succeeded where the first one didn't. Addressing a specific, high-stakes use of AI, firing people, rather than regulating digital tools broadly, gave lawmakers a much easier case to make to a governor who'd already explained precisely what he objected to the first time.

SB 947 was one of 13 AI-related bills Newsom signed the same day, forming a genuinely broad legislative package. Other bills in the group covered workplace surveillance restrictions, AI use in healthcare, limits on attorneys fully delegating legal work to AI, deepfake and digital replica rules, and updates to California's existing AI Transparency Act.

Why it matters: signing this many AI-related bills together signals California treating AI regulation as a genuinely comprehensive, ongoing legislative priority, not a single reactive measure. It follows earlier Newsom-era AI laws covering frontier AI safety and child safety protections, building out a broader regulatory framework over time rather than addressing AI concerns one isolated incident at a time.

Newsom framed the law plainly: "AI should expand opportunity, not come at the expense of workers." AFL-CIO President Lorena Gonzalez welcomed the package, saying it shows "how we can begin to put guardrails on AI at work," specifically pointing to maintaining human oversight as the core goal.

Who's affected: that framing matters because it positions the law as pro-worker rather than anti-AI, allowing companies to keep using automated tools in their hiring and management processes while drawing a clear line against letting those tools operate as the sole decision-maker in someone's employment.

Background: California's new rules follow several documented cases of AI-driven employment decisions causing real harm. A class-action lawsuit revealed that Amazon's AI-powered productivity system fired more than 300 workers at a single warehouse for failing to meet speed quotas, with bathroom and medical breaks counted against their productivity scores.

Separately, workplace software company Workday faced a collective-action lawsuit from plaintiffs over age 40 alleging its AI-driven applicant-screening tools disproportionately rejected older candidates. One plaintiff claimed the company's algorithms contributed to him being rejected from more than 100 jobs over seven years. Amazon has also faced complaints that AI-driven processes handled disability accommodation decisions in ways that didn't meet Americans with Disabilities Act standards.

Why it matters: these aren't hypothetical concerns lawmakers invented to justify new regulation, they're real, documented cases where automated systems made consequential employment decisions with limited human oversight, exactly the pattern SB 947 is built to prevent going forward.

Background: there's no comprehensive federal law specifically regulating AI's use in hiring, firing, or workplace discipline. The federal Equal Employment Opportunity Commission has issued guidance warning that AI hiring tools can run afoul of existing discrimination laws, but that's enforcement guidance applied to decades-old statutes, not a dedicated law written specifically for automated workplace decision-making the way SB 947 is.

Why it matters: that gap is part of why California acting here carries extra weight. Without federal rules specifically addressing AI in employment decisions, state-level laws like this one become the primary legal framework governing the issue, at least for now, making California's approach a likely reference point for other states considering similar legislation.

• If you work in California, your employer can no longer fire or discipline you based solely on an AI system's output, starting July 1, 2027. A human has to independently verify that decision first.
• You have a right to request a description of the data an automated system used about you, if one was involved in a decision affecting your employment.
• If you run a business using AI-driven HR or performance-management tools, build human review into termination and discipline decisions now, well ahead of the 2027 effective date.
• Watch for other states to follow California's lead. The governor's office explicitly framed several of these protections as first-in-the-nation, a pattern that's often followed by similar legislation elsewhere.