California is moving to prohibit employers from using artificial intelligence tools that read, infer or analyze workers' emotions on the job, a step that would directly restrict a category of workplace technology already embedded in some hiring and monitoring systems.

The legislative push touches people-process decisions across the employee lifecycle. Where vendors have marketed emotion-detection capabilities — in video interview scoring, call-center sentiment dashboards, productivity monitoring and engagement platforms — California employers may soon face a flat prohibition rather than a disclosure obligation. HR leaders will need to audit their existing tech stack, not just their policies, to determine whether any deployed system crosses the line.

The bill aligns with California's broader pattern of regulating workplace AI ahead of federal action. The state has already established itself as the most active US jurisdiction on algorithmic employment decisions, and an emotion-AI ban would extend that posture from hiring tools into day-to-day workforce monitoring.

What does the measure actually prohibit?

At its core, the proposal targets systems that claim to identify or infer emotional states — stress, engagement, frustration, honesty, attentiveness — from facial expressions, voice tone, keystroke patterns or other behavioral signals.

For HR teams, the practical exposure points include:

• Video interview platforms that score candidates on perceived enthusiasm or confidence
• Call-center and customer-facing tools that flag agent sentiment in real time
• Productivity and wellness monitoring that infers burnout or disengagement from digital behavior
• Meeting analytics that summarize participant engagement or mood

If enacted, employers relying on any of these capabilities in California would face a compliance deadline measured from the law's effective date, with the question shifting from "how do we disclose this?" to "can we use it at all?"

Why regulators are skeptical of emotion AI

The scientific basis for emotion recognition has drawn sustained criticism from researchers and civil liberties groups. Critics argue that facial expressions do not map reliably onto internal emotional states, and that the technology performs unevenly across demographics — raising both accuracy and discrimination concerns when outputs feed employment decisions.

That combination — contested science plus potential disparate impact — puts emotion AI in a different risk category from other workplace analytics. A tool that misreads an employee's emotional state doesn't just produce bad data; it can shape performance reviews, coaching conversations and termination decisions built on inference rather than evidence.

Employers defending such systems would also face an evidentiary burden: proving the tool measures what it claims to measure, for every population it's applied to.

What should employers do now?

Even before the bill's final language and effective date are settled, HR and procurement teams can act on the capability question: does any system we pay for infer emotion, even as a secondary feature?

A practical audit sequence looks like this:

• Inventory all AI-adjacent tools touching hiring, monitoring, engagement and performance
• Ask vendors directly whether their products perform emotion, sentiment or affect inference — and where the output flows
• Map which tools operate on California employees, since a state ban creates jurisdictional patchwork for multi-state employers
• Review contracts for indemnity and compliance representations if a vendor's capability becomes unlawful

The vendor conversation matters as much as the legal one. Suppliers often describe emotion features as "engagement scoring" or "sentiment analysis," labels that may not trigger an HR team's attention during procurement.

How does this fit the wider regulatory picture?

California is not acting in isolation. The EU's AI Act treats emotion recognition in workplaces as a prohibited high-risk application, and US states including Illinois — through its AI Video Interview Act — have already imposed notice and consent requirements on AI-mediated hiring.

A California prohibition would be a stronger intervention than disclosure regimes, and its reach across the employee lifecycle — not just candidate screening — would set the widest scope among US state measures.

For national employers, the practical consequence is fragmentation: a tool lawful in one state and prohibited in another forces either geo-fenced configuration or enterprise-wide retirement of the capability.

What happens next?

The bill still must clear the California Legislature and earn the governor's signature before any provision binds employers, and amendments during that process could narrow or clarify its scope — particularly around how regulators define emotion inference versus ordinary sentiment analytics.

HR leaders should track the final statutory language for the effective date and any enforcement mechanism, and treat the current draft as a signal to begin vendor audits now rather than wait for certainty.