AI Workforce Regulation: What Employers Need to Do Now

Artificial intelligence is no longer a future concern for employment lawyers and HR teams. It is already embedded in how many organizations recruit, manage and restructure their workforces. Regulators in California and Connecticut are responding with concrete legislative proposals and executive action that would impose new compliance obligations on employers who use these tools. Employers with UK or EU operations should also be watching parallel developments, as workplace AI is increasingly being regulated through a combination of AI-specific rules, data protection law, equality law and employment consultation obligations.

The developments described below reflect a consistent regulatory philosophy: Automated systems that affect workers must include meaningful human oversight, workers must be informed when AI is involved in decisions about them, and employers cannot hide behind algorithmic outputs when those decisions are challenged.

1. California Senate Bill 947: Human Review Before Adverse Action
   SB 947 would directly restrict how employers use automated decision systems (ADS) in disciplinary and termination decisions. Under the bill, an employer could not rely solely on ADS outputs to discipline, terminate or deactivate a worker. Where an ADS plays a primary role in the process, a human reviewer would need to independently investigate and develop corroborating evidence before any adverse action is finalized. If that review cannot substantiate the system’s output, the employer would be prohibited from proceeding.

SB 947 also includes a required notice of rights for affected workers and access rights to the data the system used. Because it is designed to operate alongside California’s existing privacy framework, employers could face layered obligations under both employment and privacy law simultaneously.

When AI plays a role in a disciplinary or termination decision, employers should ask themselves whether the process generates documentation that could independently support that decision if the algorithmic output were removed from the record entirely. If the answer is no, or even uncertain, the bill’s requirements would expose a gap.

2. California Senate Bill 951: A WARN Act for AI-Driven Workforce Changes
   SB 951 would create an advance-notice requirement modeled on traditional federal and California WARN Act concepts, but applied specifically to AI- and automation-driven workforce reductions. The trigger threshold is relatively low: 25 workers or 25 percent of the workforce affected within a 30-day period. This is a significantly lower threshold than under the current California WARN Act, which generally requires a covered employer to provide advance notice if it orders a qualifying mass layoff (when 50 or more employees suffer a separation from employment during any 30-day period due to a lack of funds or lack of work), a relocation (when all or substantially all of the employer’s operations at a covered establishment are moved 100 miles or more away), or a termination at a covered establishment (when there is the cessation or substantial cessation of industrial or commercial operations at a covered establishment).

Consistent with the California WARN Act, covered employers would need to provide 60 days’ notice to affected employees, relevant workforce agencies and local officials, along with disclosure about the automated technologies involved and available retraining resources.

The more significant feature of the bill may be its scope beyond traditional layoffs. SB 951 would also require employers to notify state authorities when they permanently stop hiring for an occupation because of AI or automation, even if no workers are actually laid off. An employer that deploys an AI system to replace future hiring in a job category, without reducing its existing headcount, could still face a notification obligation. There is no equivalent requirement under the California WARN Act.

Employers considering AI-driven operational changes should bring legal and employment counsel into the planning process before deployment decisions are made, not after. The bill, if enacted, would treat certain automation strategies as workforce actions subject to regulatory notice requirements.

3. Governor Newsom’s Executive Order N-6-26: A Preview of What Is Coming
   Executive Order N-6-26 does not create immediate employer obligations, but it is worth reading as a policy signal. The Order directs state agencies to study AI’s labor-market effects, assess potential disproportionate impacts on particular demographic groups, and develop an employment-impact dashboard. It also instructs agencies to recommend potential revisions to California’s WARN Act in light of AI-driven workforce changes and to examine mechanisms that would ensure workers share in productivity gains generated by automation, including severance practices and worker equity arrangements.

For employers, the Order suggests that California’s regulatory posture on workplace AI is still in its early stages. The measures under study now are likely to appear as legislation or agency guidance in the next few years. Organizations that build strong AI governance frameworks today will be better positioned when those requirements materialize.

4. Connecticut Senate Bill 5: Notice Before You Deploy
   Connecticut SB 5 has been enacted as Public Act 26-15, the Connecticut Artificial Intelligence Responsibility and Transparency Act (CART Act). The law imposes disclosure obligations for employers using automated employment-related decision processes, clarifies that AI use is not a defense to employment discrimination claims, and adds AI-related disclosure requirements to certain workforce reduction notices. Most provisions take effect beginning in October 2026, with key employment-AI requirements applying to covered systems deployed on or after October 1, 2027.

Organizations expanding their use of AI tools in hiring, performance management or workforce planning in Connecticut should determine whether those tools fall within the statute’s scope and whether current notice practices meet the legal standard. This is not a one-time review because as AI tools are updated or replaced, disclosure obligations should be revisited.

5. EU AI Act: High-Risk Rules for Workplace AI
   The EU AI Act is the most direct European analogue to the state-level developments described above. It classifies many AI systems used in employment, worker management and access to self-employment as high-risk, including systems used for recruitment or selection, promotion, termination, task allocation, monitoring, and evaluation of workers. For employers, that means workplace AI is not merely a general governance issue; in many cases, it will be a regulated deployment that requires risk management, transparency, appropriate documentation and effective human oversight. The EU AI Act also requires deployers of high-risk AI systems to inform workers’ representatives and affected workers before putting such systems into service or using them in the workplace.

Further, certain practices are prohibited under the EU AI Act, including AI systems used to infer emotions in the workplace. Employers using video-interview analytics, biometric tools, productivity monitoring or other systems that infer worker traits should therefore review whether the tool is permitted at all and, if permitted, whether it can be explained, overseen and challenged in practice.

The EU AI Act entered into force on August 1, 2024, and applies in phases. Prohibited practices and AI literacy obligations have applied since February 2, 2025. Following the Digital Omnibus on AI, the high-risk AI system obligations for employment-related AI (Annex III systems) will now apply from December 2, 2027. Employers should not wait for these dates to begin compliance planning; the underlying framework is now settled and early preparation is advisable.

6. EU GDPR and Platform Work Rules: Human Intervention and Algorithmic Management
   The EU’s existing data protection framework is also highly relevant. Under the GDPR, individuals have rights in relation to decisions based solely on automated processing that produce legal or similarly significant effects, including rights to obtain human review, express their point of view, and contest the decision in certain circumstances. That framework can be engaged by AI tools used to reject candidates, rank applicants, allocate work, score performance, determine pay opportunities, or support disciplinary or termination decisions.

The EU Platform Work Directive adds a more sector-specific model for algorithmic management. It applies to digital labor platforms (businesses that use algorithms or automated systems to organize work performed by individuals). This includes app-based services such as ride-hailing, food and grocery delivery, courier and logistics platforms, and on-demand task or freelance marketplaces. The Directive requires these platforms to be transparent about automated monitoring and decision-making systems and includes human oversight and review obligations for decisions such as account restriction, suspension, termination, refusal of payment and decisions affecting contractual status. It also introduces a rebuttable presumption of employment for platform workers in certain circumstances. Member States must transpose the Directive into national law by December 2, 2026. Although directed at platform work, the Directive reflects the same regulatory trend seen in the EU AI Act: Employers and platforms must be able to explain, supervise and justify algorithmic decisions that affect access to work and working conditions.

7. United Kingdom: A Fragmented but Active Regulatory Landscape
   The UK does not currently have a single AI-specific employment statute equivalent to the EU AI Act or the California and Connecticut measures discussed above. The UK position is instead developing through overlapping obligations under data protection law, equality law, employment law, health and safety duties, and regulator guidance. The Information Commissioner’s Office has made automated decision-making in recruitment and workplace monitoring a clear priority, including by scrutinizing AI-powered sourcing, screening and selection tools and emphasizing transparency, bias monitoring, meaningful safeguards and genuine human involvement.

UK data protection law is particularly important where AI tools make or meaningfully support significant decisions about workers or candidates. The Data (Use and Access) Act 2025 updates the UK automated decision-making framework and requires safeguards for significant decisions based solely on automated processing, including information for affected individuals, the ability to make representations or contest the decision, and human review. Employers should also consider whether AI-enabled monitoring, biometric attendance systems, productivity analytics or similar tools are fair, proportionate, transparent and supported by an appropriate lawful basis.

There is also no UK equivalent to an AI-specific WARN-style notice obligation for stopping hiring because of automation. However, if AI or automation leads to redundancies, existing collective consultation rules may apply where an employer proposes 20 or more redundancies at one establishment within 90 days. Equality Act risks should also be front of mind: An apparently neutral AI tool used in hiring, promotion, performance management, absence management or redundancy selection could create indirect discrimination exposure if it disadvantages people with protected characteristics and cannot be objectively justified.

How Employers Should Respond
These developments share a common thread: Regulators are no longer focused only on high-level AI governance principles. They are writing rules aimed specifically at workplace applications, and the compliance obligations they create require coordination across legal, HR, privacy and operations teams.

For most employers, the immediate priority is conducting an inventory of AI and automated decision-making tools used in employment-related contexts. Organizations should identify where AI or automated decision tools are currently used in employment contexts, assess whether adequate human oversight and documentation exist for decisions driven by those tools, and determine whether existing notice and disclosure practices would satisfy emerging statutory and regulatory requirements in the jurisdictions where they operate.

Employers evaluating new AI tools or automation-driven operational changes should treat labor, employment, privacy, and equality review as a threshold requirement, not an afterthought. The regulatory direction is clear across the United States, the UK and the EU, and the pace of legislative and regulatory activity suggests that waiting for final rules before beginning compliance work is a diminishing option.