Artificial intelligence is no longer a distant technology issue for Michigan employers. It is already part of the modern workplace. Employers use software to screen resumes, rank candidates, schedule shifts, track productivity, monitor remote work, evaluate performance, identify safety risks, draft employment documents, and assist with discipline or termination decisions. Some of these tools are marketed as artificial intelligence. Others are described more modestly as analytics, automation, workflow optimization, recruiting support, performance management, scheduling technology, or monitoring software. The label does not matter as much as the function. When a system collects worker data, scores applicants, ranks employees, recommends employment action, or influences a manager’s decision, it becomes part of the employment process and must be evaluated through the lens of employment law.
Michigan employers should now pay close attention to proposed legislation that would regulate the use of automated decision tools and electronic monitoring tools in the workplace. House Bill 5579 and Senate Bill 1077, both titled the proposed “Responsible Artificial Intelligence Security for Employees Act,” would create significant new obligations for employers that use certain AI, algorithmic, automated, or monitoring systems in employment-related decisions. ¹ ² These bills are not yet law. That distinction is important. Employers should not treat the proposed framework as a current legal mandate. But employers should treat it as a serious signal of where employment regulation is heading and as a useful roadmap for what responsible compliance will likely require.
The proposed Michigan framework reflects a broader national movement toward regulation of AI in employment. The Equal Employment Opportunity Commission has already made clear that existing federal civil rights laws apply when employers use software, algorithms, artificial intelligence, or automated systems in employment selection procedures.⁴ The EEOC and the Department of Justice have also warned that algorithmic hiring tools may violate disability discrimination laws when they screen out qualified individuals with disabilities, fail to provide reasonable accommodations, or make prohibited disability-related inquiries.⁵ ⁶ The Michigan Civil Rights Commission has similarly adopted guiding principles warning that AI systems can perpetuate discriminatory outcomes unless they are tested, limited, explained, and subject to meaningful human alternatives. ³
For Michigan businesses, especially small and mid-sized employers in Ann Arbor, Detroit, Novi, Troy, Livonia, Southfield, Grand Rapids, Lansing, and other commercial centers, the practical issue is not whether AI should be banned from the workplace. The issue is whether employers can use technology in a way that remains fair, explainable, job-related, privacy-conscious, and legally defensible. AI can help employers operate more efficiently. It can reduce administrative burdens, improve consistency, and help businesses process large volumes of information. But efficiency does not excuse discrimination, privacy invasion, retaliation, wage-and-hour violations, labor-law interference, or inaccurate employment decisions.
The proposed Michigan bills would regulate two broad categories of tools. The first category is an “automated decisions tool.” Under the proposed framework, that term would include a computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues simplified outputs, such as a score, classification, or recommendation, and is used to substantially assist or replace discretionary decision-making for employment decisions. ¹ ² That definition is important because many workplace systems do not appear dramatic or futuristic. A resume ranking system, a candidate matching score, an attendance-risk alert, a productivity dashboard, a scheduling algorithm, or a performance classification tool may all become legally significant if it substantially assists an employment-related decision.
The second major category is an “electronic monitoring tool.” The proposed bills would define that concept broadly enough to cover systems that collect data about a covered individual’s activities or communications by means other than direct observation, including computer, telephone, wire, radio, camera, electromagnetic, photoelectronic, or photo-optical systems.¹ ² This language could reach traditional workplace surveillance, but it could also reach remote-work monitoring, keystroke tracking, screen capture tools, GPS systems, badge access records, call analytics, email or chat monitoring, productivity measurements, camera systems, and other technology that tracks worker activity. Employers should therefore avoid thinking of the proposed law as only a hiring law. It is also a workplace monitoring, data governance, privacy, and employment decision-making law.
The proposed framework would apply to covered individuals, including applicants and employees. It would also reach certain independent contractors who provide services to or through an employer operating in Michigan. ¹ ² That feature matters because many modern businesses rely heavily on contractors, consultants, drivers, sales representatives, healthcare professionals, project-based workers, IT professionals, and platform-based service providers. If an employer uses automated or monitoring tools to evaluate, assign, discipline, compensate, rank, or schedule these individuals, the employer should consider whether the proposed legislation could affect those practices if enacted.
The bills would also define employment-related decisions broadly. The term would include decisions affecting wages, benefits, compensation, hours, schedules, performance evaluations, hiring, discipline, promotion, termination, job content, work assignments, access to opportunities, productivity requirements, workplace health and safety, and other terms or conditions of employment. ¹ ² This breadth is one of the most important parts of the proposed law. Employers often begin AI compliance by thinking about hiring. Hiring is important, but post-hire uses may present equal or greater risk. A tool that reduces an employee’s hours, flags an employee for discipline, ranks workers for promotion, recommends termination, evaluates safety performance, assigns less favorable work, or changes compensation can be just as consequential as a hiring screen.
One of the most striking features of the proposed Michigan legislation is that it would not merely require disclosure of AI use. It would generally prohibit employers from using automated decision tools to make employment-related decisions, subject to a limited exception for screening large volumes of job applications to identify candidates who meet set hiring criteria or to assess candidates based on job skills. ¹ ² If enacted in that form, the law would require employers to examine whether certain automated uses are permitted at all, not merely whether they are disclosed. That would be a major shift for employers that currently rely on automated systems for performance management, disciplinary recommendations, scheduling, internal mobility, productivity review, or workforce planning.
Even where use is permitted, the proposed bills would impose significant conditions. Employers using an automated decision tool or electronic monitoring tool would need to provide written notice, obtain written consent, ensure that collected data is accurate and up to date, allow covered individuals to correct inaccurate data, use the tool in a narrowly tailored manner, select the least invasive means possible, collect the least amount of data necessary, apply the tool to the smallest number of individuals necessary, and avoid collecting data when an employee is off duty. ¹ ² In practical terms, the proposal would require employers to justify why the tool is needed, why the selected data is necessary, why a less intrusive method would not work, and how the employer will prevent unnecessary collection or use.
This proposed structure is consistent with a broader principle that employers should adopt now: AI tools should be proportional to the business need. A company should not collect broad categories of employee data simply because a vendor platform makes collection easy. It should not monitor all employees constantly when limited monitoring would address a specific operational concern. It should not use a scoring system that no one understands merely because it produces fast rankings. It should not rely on a vendor’s promise of “objective” results without understanding what data the tool uses, how the tool produces outputs, and whether the outputs are valid for the job at issue.
The proposed data restrictions are also significant. The bills would prohibit certain collection of health, medical, lifestyle, wellness, protected-class, workplace-communication, device-usage, audio-video, sensor, movement, voiceprint, facial, emotion, gait, automated-tool output, online, and private social media data.¹ ² Employers should pay close attention to these categories because modern workplace technology is often designed to collect precisely these kinds of information. A remote-work tool may collect device usage and screen activity. A warehouse system may collect movement data. A call-center system may analyze tone or voice. A video-interview platform may claim to evaluate facial expression or emotional state. A scheduling platform may rely on worker availability, location, or behavioral history. A social media or background screening vendor may collect online information. Each of those practices should be carefully reviewed.
The proposed bills would also prohibit the use of covered tools to identify, punish, or obtain data about individuals engaging in activities protected by state or federal labor or employment law. ¹ ² That concern is not theoretical. The National Labor Relations Board’s General Counsel has separately warned that intrusive electronic monitoring and algorithmic management may interfere with employees’ rights to engage in protected concerted activity under federal labor law.⁸ Employers should therefore evaluate whether monitoring tools could chill protected activity, reveal organizing efforts, identify employees discussing workplace conditions, or create the appearance that employees are being watched for engaging in legally protected conduct.
The proposed privacy restrictions would extend to physical spaces as well. The bills would prohibit monitoring bathrooms and similar private areas, including locker rooms, changing areas, breakrooms, smoking areas, employee cafeterias, lounges, areas designated for expressing breast milk, and areas designated for prayer or religious activity. ¹ ² The proposal would also restrict certain monitoring of workplace activity in an employee’s residence, personal vehicle, or property owned or leased by the employee. ¹ ² These provisions are especially important in the remote-work era. Employers may view remote-work monitoring as a software issue, but employees may experience it as a direct intrusion into the home. Employers should treat remote-work surveillance as a high-risk practice that requires a strong justification, narrow scope, clear disclosure, and careful limits.
Another important feature is the proposed prohibition on covered tools equipped with facial, gait, voice, or emotion recognition technology. ¹ ² This language targets tools that attempt to infer workplace qualities from physical, behavioral, or expressive traits. Employers should be cautious about products that claim to evaluate honesty, attention, enthusiasm, emotional stability, reliability, or cultural fit based on video interviews, voice tone, facial movement, eye contact, gait, or similar characteristics. These systems may be difficult to validate, difficult to explain, and vulnerable to discrimination claims involving disability, race, national origin, age, sex, accent, or other protected characteristics. Even outside the proposed Michigan legislation, the legal risk is substantial when a tool relies on traits that may not be job-related or may operate as proxies for protected status.
Notice would be a central requirement under the proposed Michigan framework. Employers would need to display a workplace poster regarding use of electronic monitoring or automated decision tools. They would need to provide written notice before implementing a covered tool, include notice in job postings, post notice on the employer’s website, provide notice directly to applicants, and make notice available in accessible formats that account for disability and language needs. ¹ ² This would require coordination among human resources, recruiting, legal, management, website administrators, and vendors. A vague sentence in an employee handbook would likely not be enough. Employers would need clear, specific, and accessible explanations of what tool is used, what data is collected, why it is collected, what employment decisions may be affected, and what rights the applicant or employee has.
The proposed opt-out requirement may be one of the most operationally difficult parts of the bills. The notice would need to give covered individuals the ability to opt out of the electronic monitoring or automated decision tool. If the individual opts out, the employer could not use the tool to make employment-related decisions for that person. ¹ ² That would require employers to maintain a parallel human process. An employer could not simply state that the company uses AI and then force every applicant or employee into the system. If enacted, the law would require a practical alternative for individuals who decline the automated or monitoring process. This would require planning, staffing, documentation, and training.
Impact assessments would be another major compliance obligation. Before using a covered automated decision tool or electronic monitoring tool, an employer would need to conduct an impact assessment evaluating the tool’s objectives, algorithms, data, cybersecurity vulnerabilities, and potential biases, including discriminatory outcomes based on race, gender, or disability. ¹ ² The assessment would need to be conducted by an independent and impartial third party with no financial or legal conflicts of interest related to use of the tool. ¹ ² The assessment would also need to evaluate whether the tool’s attributes and modeling techniques are scientifically valid for assessing performance or ability to perform essential job functions, whether the tool’s attributes may function as proxies for protected classes, whether training data or outputs create disparate impact, whether accessibility is limited for individuals with disabilities, and whether less discriminatory alternatives exist. ¹ ²
This impact-assessment requirement would create several practical challenges. First, employers would need access to technical information about the tool. Many vendors treat algorithms, training data, model logic, validation studies, and scoring methods as proprietary. If a vendor cannot provide enough information for an independent assessment, the employer may not be able to use the tool safely. Second, employers would need interdisciplinary review. AI employment risk involves employment law, civil rights law, privacy, cybersecurity, data governance, statistics, disability accommodation, labor relations, and operational management. Third, the assessment would not be a one-time exercise. The proposed bills would require periodic review, including annual reassessment while the tool remains in use. ¹ ²
The proposed public registry requirement would add another layer of consequence. The bills would require employers to submit the assessment, either in full or in accessible summary form, to the Michigan Department of Labor and Economic Opportunity for inclusion in a public registry of impact assessments. ¹ ² Employers would also need to distribute the assessment to covered individuals who may be subject to the tool. ¹ ² This is a major transparency concept. Employers are accustomed to keeping internal audits, vendor evaluations, legal risk assessments, and HR compliance materials confidential. A public or employee-accessible assessment would need to be carefully written, accurate, complete, and consistent with actual practice. Statements made in an assessment could later matter in litigation, agency proceedings, union negotiations, public relations disputes, or vendor disputes.
The proposed documentation obligations would also affect vendor contracting. Employers would need to retain documentation concerning the design, development, use, and data of the tool that may be necessary to conduct an impact assessment. That documentation would include the source of data used to develop the tool, technical specifications, individuals involved in development, historical use data, and historical records of tool versions used by the employer. ¹ ² Service providers contracting with employers would need to allow access to that documentation. ¹ ² Employers should therefore review vendor contracts now. AI procurement contracts should address audit rights, documentation access, validation materials, model changes, data use restrictions, cooperation with impact assessments, breach notice, indemnification, retention obligations, and termination rights if the vendor cannot support compliance.
Data retention and data sharing would be regulated under the proposed framework. Employers collecting covered data would need to retain it for not more than three years after the purpose for using the tool is achieved, unless a collective bargaining agreement provides otherwise. If the employer does not use specific data, it would need to delete that data immediately. ¹ ² The proposal would also prohibit selling or licensing covered individual data, including deidentified or aggregated data, and would restrict sharing such data with state or local government except for specified purposes. ¹ ² This reflects a broader shift from data accumulation to data minimization. Employers should not collect employment data simply because it may someday be useful. They should collect only what is necessary, use it only for the stated purpose, protect it carefully, and delete it when no longer needed.
The proposed breach-response obligations would be unusually serious. If an employer experiences a security breach involving data collected through an electronic monitoring tool or automated decision tool, the employer would need to secure systems, mitigate harm, certify corrective steps, notify affected individuals within a short period after discovery, notify state authorities, and provide significant identity-theft and financial-monitoring protections. ¹ ² Whether these provisions change during the legislative process remains to be seen, but their inclusion shows that lawmakers view AI and monitoring data as sensitive. Employers should treat cybersecurity as part of employment AI compliance, not as a separate IT function. A weak security program can turn a questionable AI practice into a major employment, privacy, and reputational crisis.
The proposed Michigan legislation also has labor-relations implications. The bills would provide minimum standards and would not preempt or diminish employees’ rights to collectively bargain over terms and conditions of employment, including protections against surveillance-based wage discrimination. ¹ ² If employees are covered by a collective bargaining agreement and the employer intends to use covered tools to set or influence wages or other terms and conditions of employment, the employer would need to provide notice and an opportunity to bargain. ¹ ² Unionized employers should therefore be cautious before implementing AI-based scheduling, productivity, performance, discipline, compensation, or monitoring systems without evaluating bargaining obligations and protected activity concerns.
The enforcement provisions would create litigation risk. A covered individual aggrieved by a violation, or a person acting on behalf of such an individual, including a labor organization, could bring an action for damages, injunctive relief, or both. ¹ ² A prevailing plaintiff could recover economic damages, noneconomic damages, costs, and attorney fees. ¹ ² The proposed bills would also authorize civil fines and enforcement by public authorities. ¹ ² For employers, the attorney-fee and injunctive-relief provisions may be especially important. Even where individual damages are modest, a fee-shifting statute can make AI compliance disputes attractive to plaintiffs, unions, and advocacy organizations.
Employers should also remember that proposed AI-specific legislation is not the only risk. Existing law already applies. Title VII prohibits employment practices that discriminate based on race, color, religion, sex, or national origin, and the EEOC has explained that automated systems used in employment selection procedures may create adverse impact liability if they disproportionately exclude protected groups and are not properly justified.⁴ The Uniform Guidelines on Employee Selection Procedures remain important because they provide the traditional framework for evaluating selection procedures, adverse impact, and validation.⁷ An AI hiring tool may look new, but the legal questions are familiar: What does it measure? Is it job-related? Is it valid? Does it disproportionately screen out protected groups? Is there a less discriminatory alternative?
Disability discrimination presents a separate and significant risk. The EEOC and DOJ have warned that AI and algorithmic tools may violate disability laws when they screen out qualified individuals, fail to provide reasonable accommodations, or require disability-related information in ways that are not legally permitted.⁵ ⁶ Employers should be especially careful with tools that measure response speed, facial expression, speech patterns, keyboard use, memory, personality traits, physical movement, game performance, or video interview behavior. A tool may appear neutral while disadvantaging applicants or employees with disabilities. Employers should provide reasonable accommodations, offer alternative formats where appropriate, train HR personnel to recognize accommodation requests, and avoid tools that cannot be adapted without undermining their claimed validity.
The Michigan Civil Rights Commission’s AI principles reinforce these concerns at the state level. The Commission has recognized the risk that AI systems can produce discriminatory outcomes if fundamental principles and objective guidelines are not adopted. ³ The Commission’s principles emphasize protection from unsafe or ineffective systems, prevention of algorithmic discrimination, privacy protections, notice and explanation, human alternatives, regular assessments, and protections for hiring, retention, recruitment, and promotion. ³ Even though those principles are not the same as enacted employment legislation, they are important because they show how Michigan civil rights authorities may approach AI-related harms.
The National Institute of Standards and Technology’s AI Risk Management Framework provides a useful governance model for employers even though it is voluntary and not employment-specific. The framework emphasizes mapping, measuring, managing, and governing AI risks.⁹ For employers, those concepts translate into practical steps: identify where AI is used, understand what data it relies on, measure potential bias or inaccuracy, assign internal responsibility, document decisions, monitor performance, and update controls as the tool changes. Employers do not need to become software engineers, but they do need a governance structure that prevents AI from being adopted casually without legal, operational, and human review.
New York City’s automated employment decision tool law also provides a useful comparison. Local Law 144 requires certain automated employment decision tools used in hiring and promotion to undergo a bias audit, requires public availability of audit information, and requires notice to employees or job candidates. ¹⁰ Michigan’s proposed framework is broader in several respects because it would address electronic monitoring, post-hire employment decisions, opt-out rights, extensive data restrictions, impact assessments, retention, breach obligations, and labor-related concerns. But New York’s experience shows that AI employment regulation is not hypothetical. Employers operating across jurisdictions may soon face a patchwork of federal, state, and local rules.
The first practical step for Michigan employers is to conduct an AI and electronic monitoring inventory. Many employers do not know how many automated systems they already use. Human resources may use one applicant tracking system. Payroll may use analytics. Operations may use productivity dashboards. IT may use monitoring software. Security may use cameras, access logs, or facial recognition tools. Supervisors may use generative AI informally to draft evaluations, discipline documents, or interview questions. The inventory should identify each tool, vendor, internal owner, data collected, output generated, employment decision affected, individuals covered, human review process, retention period, and contract terms.
The second step is to classify tools by risk. Not every AI or automated system presents the same level of employment risk. A tool that helps summarize general training materials is different from a tool that ranks applicants for interviews. A scheduling aid that displays availability is different from an algorithm that allocates hours based on productivity scores or customer ratings. A grammar assistant is different from a system that drafts disciplinary warnings based on employee monitoring data. Employers should prioritize tools that affect hiring, promotion, discipline, termination, compensation, scheduling, performance evaluation, work assignments, safety, accommodations, or access to opportunities.
The third step is to review whether each tool is job-related, valid, and explainable. Employers should be able to answer basic questions before relying on a tool in any meaningful employment decision. What does the tool measure? Why does that measurement matter to the job? What data does the tool use? Is the data accurate? Has the tool been validated? Does the tool produce disparate impact? Is it accessible to individuals with disabilities? Can a human reviewer override the output? Can the employer explain the final decision without hiding behind the vendor? If the employer cannot answer those questions, the tool should not be used for consequential employment decisions until additional review is completed.
The fourth step is to strengthen vendor contracts. Many AI and HR technology contracts are written like ordinary software agreements. That is not enough for employment AI tools. Employers should require vendors to disclose whether the product uses AI, machine learning, statistical modeling, automated scoring, ranking, classification, recommendations, or monitoring. Contracts should require cooperation with audits and impact assessments, access to relevant technical documentation, notice of model changes, data deletion rights, cybersecurity protections, breach notice, restrictions on vendor use of employee data, and indemnification for vendor-caused legal violations. Employers should also avoid contract terms that leave all employment-law responsibility with the employer while denying the employer access to the information needed to evaluate the tool.
The fifth step is to preserve meaningful human review. Human review should not be symbolic. If a manager automatically accepts a system recommendation, the decision may still be effectively automated. Meaningful review requires training, authority, independent judgment, and documentation. The human reviewer should understand the tool’s limitations, check the underlying facts, consider accommodations or exceptions, evaluate whether the result makes sense, and document the legitimate business reason for the decision. Employers should also consider appeal or reconsideration procedures when applicants or employees believe an automated or monitoring tool produced an inaccurate or unfair result.
The sixth step is to address accommodations and accessibility. Employers should not wait for an applicant or employee to be harmed before considering whether a tool is accessible. If a video interview tool, game-based assessment, timed test, voice analysis system, physical movement tracker, or online platform disadvantages certain individuals with disabilities, the employer may face legal exposure. Employers should provide notice of assessment methods where appropriate, offer alternative procedures, train HR personnel, document accommodation options, and avoid tools that cannot be reasonably modified.
The seventh step is to adopt a generative AI policy for employment use. Many risks arise outside formal vendor systems. A supervisor may paste employee notes into a public AI tool and ask it to draft a termination memo. A recruiter may ask AI to rank resumes. A manager may use AI to summarize an employee complaint. A business owner may ask AI to compare two employees for layoff selection. These informal uses can disclose confidential information, create inaccurate records, introduce bias, or generate language that becomes damaging evidence. Employers should define when generative AI may be used in HR, what information may not be entered, who must approve use, and what human review is required.
The eighth step is to improve notice and transparency. Even before Michigan enacts any AI-specific law, employers should consider providing clear explanations when technology affects applicants or employees. Transparency reduces surprise, improves trust, and forces the employer to understand its own systems. A useful notice should identify the tool, describe the data collected, explain the purpose, identify the employment decisions affected, state whether human review occurs, explain how inaccuracies can be corrected, describe accommodation options, identify retention practices, and provide a human contact for questions.
The ninth step is to minimize data collection. Employers should stop treating employee data as something to collect first and justify later. AI and monitoring tools can generate enormous amounts of information. Excess data creates privacy risk, cybersecurity risk, discovery risk, and employee-relations risk. Employers should identify what data is necessary, what data is merely convenient, what data is sensitive, what data should never be collected, and when data should be deleted. Data collected for cybersecurity should not automatically become performance data. Data collected for safety should not automatically become discipline data. Data collected for scheduling should not automatically become promotion data. Purpose limitation should become a core employment AI principle.
The tenth step is to train managers, HR personnel, executives, and IT staff. AI compliance is not only a legal department issue. Managers need to understand that an AI recommendation is not a legal defense. HR needs to understand disparate impact, accommodations, and documentation. IT needs to understand retention, access controls, and cybersecurity. Executives need to understand that technology procurement can create employment-law exposure. Legal counsel needs to understand enough about the technology to ask meaningful questions. Training should focus on real workplace scenarios, including applicant screening, remote monitoring, productivity scoring, discipline, scheduling, accommodations, employee complaints, protected activity, and data breaches.
Small employers should not assume this issue only affects large corporations. The proposed Michigan framework would reach employers operating in Michigan and would also account for third parties and service providers used in employment-related decisions. ¹ ² Small and mid-sized businesses may actually be more vulnerable because they often rely on off-the-shelf platforms without internal legal, statistical, or technical review. A small employer using an inexpensive applicant tracking system, payroll analytics tool, remote monitoring platform, or scheduling algorithm may not realize that the system is producing legally significant scores, rankings, or recommendations. The practical solution is not to avoid technology altogether, but to choose tools carefully, limit their use, and document the reasons for relying on them.
Employers should also consider the human impact. AI monitoring and automated decision tools can damage workplace trust when employees believe they are being watched constantly, judged by inaccurate data, or disciplined by systems no one can explain. A legally compliant system can still create morale problems if employees view it as unfair or dehumanizing. Employers should communicate the business reason for technology, limit monitoring to legitimate needs, avoid surprise implementation, invite questions, provide human contacts, and correct errors quickly. The more invasive the tool, the stronger the justification should be.
The proposed Michigan legislation may change. It may be amended, narrowed, expanded, delayed, or rejected. The final law, if enacted, may differ substantially from the introduced bills. But employers should not wait for final enactment to begin preparation. Most of the preparation steps are good compliance practices regardless of whether the current proposal becomes law. Inventorying AI tools, evaluating bias, improving contracts, documenting job-relatedness, protecting employee data, maintaining human review, providing accommodations, and limiting monitoring will help employers comply with existing civil rights and employment laws.
The best way to think about AI employment compliance is to treat the technology as part of the decision-maker. If an employer would need fairness, documentation, job-relatedness, consistency, confidentiality, and accountability from a human decision-maker, it should expect at least the same scrutiny when technology assists the decision. AI may make employment decisions faster, but speed is not a substitute for legality. A faster discriminatory process is still discriminatory. A faster inaccurate process is still unfair. A faster opaque process may be harder to defend.
Michigan employers should begin now by asking a simple question: could we explain this tool to an applicant, employee, union representative, agency investigator, judge, jury, or news reporter? If the answer is no, the employer should slow down. The coming wave of AI employment regulation will reward employers that understand their systems, limit data collection, document legitimate business reasons, protect employee privacy, test for discrimination, and preserve meaningful human judgment. It will punish employers that adopt tools they cannot explain, rely on vendors they have not vetted, collect data they do not need, and use automated outputs as a substitute for managerial responsibility.
The proposed Michigan AI employment rules should therefore be viewed as a practical compliance roadmap. They identify the issues that lawmakers, agencies, employees, unions, and courts are likely to scrutinize notice, consent, data accuracy, privacy, protected-class bias, accessibility, impact assessments, vendor accountability, cybersecurity, human alternatives, labor rights, and remedies. Employers that prepare for those issues now will be better positioned whether the bills pass in their current form, pass in amended form, or simply influence the standard of care expected of responsible Michigan businesses.
This article is for general educational and informational purposes only. It is not legal advice and does not create an attorney-client relationship. Employers considering the use of artificial intelligence, automated decision tools, or employee monitoring systems should consult counsel regarding their specific workforce, technology, contracts, policies, industry, and legal obligations.
Contact Tishkoff
Tishkoff PLC specializes in business law and litigation. For inquiries, contact us at www.tish.law/contact/. & check out Tishkoff PLC’s Website (www.Tish.Law/), eBooks (www.Tish.Law/e-books), Blogs (www.Tish.Law/blog) and References (www.Tish.Law/resources).
References and Citations
1- Michigan House Bill 5579, 103rd Legislature, 2025–2026 Regular Session, introduced February 24, 2026, proposed “Responsible Artificial Intelligence Security for Employees Act.” https://legislature.mi.gov/documents/2025-2026/billintroduced/House/htm/2026-HIB-5579.htm
2- Michigan Senate Bill 1077, 103rd Legislature, 2025–2026 Regular Session, introduced June 24, 2026, proposed “Responsible Artificial Intelligence Security for Employees Act.” https://legislature.mi.gov/Bills/Bill?ObjectName=2026-SB-1077
3- Michigan Civil Rights Commission, “MCRC Guiding Principles for the Elimination and Prevention of Artificial Intelligence Bias and Discrimination,” adopted October 2024. https://www.michigan.gov/mdcr/news/releases/2024/10/21/mcrc-passes-resolution-to-establish-guiding-principles-for-use-of-ai-in-michigan
4- U.S. Equal Employment Opportunity Commission, “Select Issues: Assessing Adverse Impact in Software, Algorithms, and Artificial Intelligence Used in Employment Selection Procedures Under Title VII of the Civil Rights Act of 1964,” EEOC-NVTA-2023-2, issued May 18, 2023. https://www.theemployerreport.com/2023/05/eeocs-new-guidance-focuses-on-adverse-impact-in-ai-used-in-employment-selection-procedures/
5- U.S. Equal Employment Opportunity Commission, “The Americans with Disabilities Act and the Use of Software, Algorithms, and Artificial Intelligence to Assess Job Applicants and Employees,” issued May 12, 2022. https://www.eeoc.gov/eeoc-disability-related-resources/artificial-intelligence-and-ada
6- U.S. Department of Justice, Civil Rights Division, “Algorithms, Artificial Intelligence, and Disability Discrimination in Hiring,” issued May 12, 2022. https://www.justice.gov/archives/crt/ai
7- Equal Employment Opportunity Commission, Civil Service Commission, Department of Labor, and Department of Justice, “Uniform Guidelines on Employee Selection Procedures,” 29 C.F.R. Part 1607. https://www.eeoc.gov/laws/guidance/questions-and-answers-clarify-and-provide-common-interpretation-uniform-guidelines
8- National Labor Relations Board, Office of the General Counsel, Memorandum GC 23-02, “Electronic Monitoring and Algorithmic Management of Employees Interfering with the Exercise of Section 7 Rights,” issued October 31, 2022. https://www.laborrelationslawinsider.com/2024/11/nlrb-joins-regulatory-assault-on-electronic-surveillance-of-the-workplace/
9- National Institute of Standards and Technology, “Artificial Intelligence Risk Management Framework (AI RMF 1.0),” NIST AI 100-1, published January 26, 2023. https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf
10- New York City Department of Consumer and Worker Protection, “Automated Employment Decision Tools,” Local Law 144 of 2021 and related enforcement materials. https://www.osc.ny.gov/state-agencies/audits/2025/12/02/enforcement-local-law-144-automated-employment-decision-tools