California Bans AI Emotional Surveillance at Work — What Global Employers Need to Know

On 11 September 2026, California Governor Gavin Newsom signed Assembly Bill 1883 into law, prohibiting employers from using AI-powered surveillance tools to collect neural data or identify a worker's emotional state. The law is a significant step in a broader regulatory shift that is beginning to draw hard limits around how artificial intelligence can be deployed against employees — and professional services businesses operating across multiple jurisdictions would be unwise to treat this as a California-only concern.

What AB 1883 Actually Prohibits

The legislation targets two specific categories of AI use in the workplace. First, it bans the collection of neural data — information derived from brain activity or neurological signals — by employer-deployed AI tools. Second, and with arguably wider practical reach, it prohibits the use of AI systems designed to detect or infer a worker's emotional state.

That second category covers a surprisingly broad range of tools already in commercial use: sentiment analysis software that parses employee communications, video call platforms that score facial expressions during meetings, productivity monitoring tools that claim to flag disengagement, and AI systems marketed as detecting stress or deception during performance reviews. If the tool is using AI to draw inferences about how a worker feels, AB 1883 puts it in scope.

Employers based outside California are not automatically exempt. The law applies to workers in California regardless of where the employer is headquartered. A London-based HR consultancy managing remote staff in San Francisco, a Toronto accounting firm with a West Coast practice group, or a Dubai marketing agency employing contractors in Los Angeles all face potential exposure.

Why This Matters Beyond California

California has a well-documented history of setting standards that other jurisdictions eventually adopt or use as a reference point. But this time, employers do not need to wait for the diffusion effect — equivalent regulatory pressure already exists in several markets simultaneously.

In the European Union, the AI Act classifies systems that evaluate or infer the emotional states of individuals in workplaces as high-risk AI. High-risk classification triggers mandatory conformity assessments, transparency obligations, and strict limits on deployment. EU-based professional services firms, and any firm processing the data of EU workers, are already operating under a framework that treats emotional inference tools with significant scepticism.

The UK's Information Commissioner's Office has issued clear guidance that processing sensitive personal data — which inferred emotional states would likely constitute — requires explicit legal basis and a Data Protection Impact Assessment. The ICO has not been passive in enforcement.

In Canada, Bill C-27 and its proposed Artificial Intelligence and Data Act (AIDA) signal a federal intent to regulate high-impact AI systems, with emotional inference in employment contexts likely to attract scrutiny. Several Canadian provinces have also updated or are updating private sector privacy legislation.

Across the Middle East, regulators in the UAE and Saudi Arabia are developing AI governance frameworks that draw heavily on EU precedent. Asia-Pacific jurisdictions including Singapore, Australia, and Japan are each at different stages of AI-specific rulemaking, but the direction of travel — greater accountability, restricted use of sensitive inference tools — is consistent.

The practical consequence for professional services businesses is this: if you are operating across multiple jurisdictions and using AI tools that touch employee monitoring or sentiment analysis, you are almost certainly facing compliance obligations right now, not at some future point.

The Legal Risk Is Already Materialising

The same week AB 1883 was signed, a separate story underlined how quickly AI-related legal exposure is becoming real and personal. The New Mexico Supreme Court fined lawyer Stephen Aarons and held him in contempt after he submitted a brief in a murder appeal containing fabricated witness testimony and facts generated by OpenAI's ChatGPT. The court's response was unambiguous: reliance on AI-generated content without verification is not a mitigating circumstance — it is the problem.

For law firms, this is a direct professional risk. For accountants, HR consultancies, and other professional services businesses that rely on the quality and accuracy of their outputs, the principle extends equally. AI tools require governance. Outputs require verification. Accountability sits with the professional, not the model.

These two stories from the same week — one about what AI can do to employees, one about what AI can do to clients — reflect the twin compliance challenges now facing the sector.

What Employers Should Do Now

Audit your AI tools against a current regulatory map. Many businesses have adopted workplace monitoring or productivity tools without a clear understanding of what data those tools collect or infer. Vendor marketing language is not a compliance basis. You need to know, precisely, whether any tool in your stack analyses facial expressions, voice tone, communication sentiment, or physiological signals.

Assess jurisdictional exposure, not just your home market. If you have employees, contractors, or managed staff in California, the EU, the UK, or Canada, your obligations are layered. A single global policy based on your most permissive jurisdiction is a liability.

Review your employment contracts and staff communications policies. Many AI monitoring tools have been deployed without adequate disclosure to workers. Depending on jurisdiction, this may already constitute a breach — of data protection law, employment law, or both.

Establish a governance framework for AI in HR processes. This means documented decision-making on which tools are permitted, how they are assessed, who is responsible for monitoring compliance, and how workers can raise concerns. In the EU this is mandatory for high-risk AI; in most other markets it is rapidly becoming the expected standard.

Do not rely on AI output without human review. Whether you are generating client-facing documents, legal submissions, financial summaries, or HR assessments, AI output must be verified by a qualified professional before it carries your firm's authority. The New Mexico contempt ruling is not an isolated incident — it is a signal of where courts and regulators are heading.

The Cost of Inaction

Professional services businesses are trusted precisely because they are held to higher standards. When AI-related compliance failures occur in this sector, the consequences tend to be acute: regulatory fines, professional sanctions, reputational damage, and — as the New Mexico case demonstrates — personal liability for individual practitioners.

The regulatory environment around AI is not stabilising. It is accelerating. AB 1883 is one data point in a pattern that spans California, Brussels, London, Ottawa, Dubai, and Singapore. Businesses that treat each new law as an isolated local issue will find themselves perpetually reactive.

If your firm uses AI tools in employment, HR, or client-facing work and you are not certain of your compliance position across your operating jurisdictions, that uncertainty is itself a risk.

Ops Intel helps professional services businesses understand and meet their AI compliance obligations — across every market they operate in. Visit https://www.opsintel.io to learn how we can help your firm get ahead of the regulatory curve.