New technologies, including tools driven by artificial intelligence (AI), are increasingly being used in the workplace for a wide range of purposes such as measuring employee productivity, preventing theft, and monitoring workers. These technologies offer to many companies potential solutions that help optimize efficiencies and support operations, reduce human bias, prevent discrimination and harassment, and improve worker health and safety. However, the use of these tools simultaneously raises concerns if employers use these tools for anti-union purposes such as screening out candidates who are affiliated with a union. At the same time, many unions are increasingly concerned with massive job displacement because AI can potentially disrupt many occupations and even entire industries or sectors, especially with the advent of generative AI tools like ChatGPT and self-driving vehicle technologies.

This Article discusses the various ways the government has prioritized combatting AI and other emerging technologies if they are used for anti-union purposes or displace workers either intentionally or unintentionally. In doing so, this Article examines the role that executive orders, interagency agreements, specific agency actions, and regulatory proposals play in addressing AI being used for anti-union purposes or to displace workers. This Article then outlines the legislative proposals to address these issues as well as the new collective bargaining strategies and private initiatives that unions are actively pursuing to address AI-related risks. Finally, this Article concludes that it is imperative that employers, unions, and others collaborate with the government to facilitate shared goals. 

Introduction

Systems and tools using artificial intelligence (AI) have revolutionized the workplace.¹1.See generally Bradford J. Kelley, All Along the New Watchtower: Artificial Intelligence, Workplace Monitoring, Automation, and the National Labor Relations Act, 107 Marq. L. Rev. 195 (2023) (hereinafter Kelley, All Along the New Watchtower). Many of these new technologies—including wearable devices, security cameras, GPS tracking devices and cameras that keep track of the productivity and location of employees, and computer software that takes screenshots or audio recordings—enhance employers’ abilities to monitor and surveil the workplace.²2.See Bradford J. Kelley, Wage Against the Machine: Artificial Intelligence and the Fair Labor Standards Act, 34 Stan. L. & Pol’y Rev. 261, 266 (2023). For the purposes of this Article, AI refers to computer systems and algorithms utilized in a work environment to perform tasks that typically require human-level intelligence to optimize aspects of the workplace, including enhancing productivity, streamlining operations, and improving decision-making. Id. Employers are increasingly using algorithmic-driven analysis and predictive software to determine who gets interviewed, hired, promoted, developed, and disciplined. When AI is appropriately designed and administered, AI and algorithms have been shown to reduce subjectivity in employment decisions, improve diversity, encourage fairness, and make workplaces safer and more accessible.³3.See generally Keith E. Sonderling et al., The Promise and the Peril: Artificial Intelligence and Employment Discrimination, 77 U. Mia. L. Rev. 1 (2022). Labor unions are increasingly encouraging workers to use AI-powered chatbots that answer questions about workplace policies, help connect workers, and generate insights about worker experiences.⁴4.See WorkIt Labs, Our Tech: Chatbots & Gen AI, https://www.workitlabs.org/chatbot-genai (https://perma.cc/B8WK-5X4D) (last visited June 7, 2024). See also Caroline O’Donovan, The Future Of Organized Labor Could Be This Artificially Intelligent Bot, BuzzFeed.News (Nov. 14, 2016, 1:41 PM), https://www.buzzfeednews.com/article/carolineodonovan/the-future-of-organized-labor-could-be-this-artificially-int (https://perma.cc/A86L-DALC) (explaining the importance of AI tools for labor unions).

Despite the benefits of this technology, there are growing concerns that these tools could be used by employers for anti-union purposes. Critics argue that employers engaging in surveillance, or giving the impression of surveillance, can violate the National Labor Relations Act (NLRA) if their conduct has a chilling effect on protected activities and makes employees fearful of retaliation.⁵5.National Labor Relations Act, 29 U.S.C. §§ 151-59 (2018). The NLRA is the primary federal statute governing collective employee action including unionization and collective bargaining. For example, critics point to certain tracking devices that can give an employer information about the locations and times workers gather. Other concerns focus on keystroke software that could be used to identify workers’ use of specific words or phrases such as “union.” In a similar vein, critics allege that AI tools could be used by employers to screen out candidates who are (or were) affiliated with a union.

Many unions and their supporters are simultaneously concerned with massive job displacement because AI has the potential to dramatically transform the workforce. Some unions have become increasingly concerned that AI and robots will replace certain jobs, ranging from bartenders to security guards.⁶6.See Yossi Sheffi, The UAW and Other Unions Must Focus More on AI and Automation in Their Negotiations, Harv. Bus. Rev. (Sept. 15, 2023), https://hbr.org/2023/09/the-uaw-and-other-unions-must-focus-more-on-ai-and-automation-in-their-negotiations (https://perma.cc/GD4Q-7NXY). Similarly, several unions have voiced concerns that algorithms could potentially take over much of the work of product assembly lines or even entire industries.⁷7.See Tom McGrath, Liz Shuler Wants AI to Reinvigorate the Labor Movement, Politico (Mar. 31, 2024, 7:00 AM ET), https://www.politico.com/news/magazine/2024/03/31/ai-labor-power-schuler-00144086 (https://perma.cc/53TH-MCSU). For instance, there have been growing concerns within the entertainment industry that generative AI tools such as ChatGPT could replace the writers or possibly relegate them to simply refining AI-generated scripts. There are related concerns that AI-generated simulations of actors could replace the actors themselves.⁸8.Sheffi, supra note 6. At the same time, labor union leaders have stated that they believe AI-related risks can be an effective way to reinvigorate a stagnant labor movement.⁹9.McGrath, supra note 7 (discussing how the president of one of America’s largest labor organizations “sees changing technology as a great risk — and great opportunity.”).

The risk of AI being used by employers for anti-union purposes or to displace workers on a massive scale has led to widespread regulatory responses across the federal government to implement greater oversight, prevent the misuse of AI, and protect jobs. As a candidate in 2020, President Joseph R. Biden Jr. regularly stated that he would be the most pro-union president in history and would make increased unionization a top priority of the current administration.¹⁰10. Will Weissert & Joey Cappelletti, Biden says he’s the ‘most pro-union president in U.S. history’. A massive looming auto-workers strike would put that to the test, Fortune (Sept. 13, 2023), https://fortune.com/2023/09/13/biden-most-pro-union-president-us-history-massive-looming-uaw-auto-workers-strike-put-to-test (https://perma.cc/QW9T-8P4K). As a result, the Biden administration has prioritized addressing AI if it is being used for anti-union purposes or to displace workers intentionally or unintentionally.¹¹11. See infra Part I. This pro-union approach to regulating AI has involved executive orders, interagency agreements, and individual agency actions.

Executive orders have played the most prominent role in this union-protective approach to regulating AI. In 2023, President Biden issued a lengthy and far-reaching executive order regarding AI which reflects the pro-union focus of the administration.¹²12. Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, Exec. Order No. 14,110, 88 Fed. Reg. 75,191 (Nov. 1, 2023). The order gave broad directives to a wide variety of federal agencies to address AI, and it required agencies to include workers and labor unions in the decision-making process.¹³13. Id. at 75,191-192, 75,210. Another key element of the federal government’s approach to regulating the misuse of AI is interagency agreements to enhance information sharing, investigations, enforcement, training, and outreach.¹⁴14. See Kelley, All Along the New Watchtower, supra note 1, at 199. In addition, the federal government’s approach to regulate anti-union AI use has relied on specific federal agency action. The National Labor Relations Board (NLRB), the independent agency responsible for enforcing the NLRA, is the most aggressive federal agency seeking to regulate AI used for anti-union purposes. Most significantly, in 2022, the NLRB’s general counsel released a memorandum warning employers that using electronic surveillance and automated management technologies presumptively violates employee rights under the NLRA.¹⁵15. Memorandum GC 23-02 from Jennifer A. Abruzzo, Gen. Couns., Nat’l Lab. Relations Bd., to All Regional Directors, Officers-in-Charge, and Resident Officers, at 1, 8 (Oct. 31, 2022) (available at https://www.nlrb.gov/news-outreach/news-story/nlrb-general-counsel-issues-memo-on-unlawful-electronic-surveillance-and (https://perma.cc/62N2-GZB7)) (hereinafter Memorandum GC 23-02). See also Kelley, All Along the New Watchtower, supra note 1, at 215-20.

Fears over massive worker displacement caused by emerging technologies and concerns with AI being used to target or otherwise minimize the influence of unions have led to widespread legislative proposals at the federal and state levels.¹⁶16. See Zach Williams, Unions Keep Pressure on Statehouses to Regulate AI, Protect Jobs, Bloomberg L. (Sept. 18, 2023, 4:00 AM CT), https://news.bloomberglaw.com/artificial-intelligence/unions-keep-pressure-on-statehouses-to-regulate-ai-protect-jobs (https://perma.cc/JQ2L-6U92). In recent years, unions have supported legislative actions in Congress and at the state level to limit how employers use AI, especially tools that monitor and discipline employees. Legislators and unions have focused some of their legislative efforts on specific industries, such as the entertainment industry.¹⁷17. Id.

Even without national or state-level regulation, AI’s increasing ubiquity and expanding commercial potential has led some unions to turn to self-regulation and collective bargaining to address risks associated with AI. Some unions have attempted to add entirely new protections into their labor contracts, while other unions have focused on updating existing language to account for AI and other emerging technologies. Two of the most high-profile examples have come out of Hollywood where both the Writers Guild of America and SAG-AFTRA, which represents actors and writers, went on monthslong strikes after their labor contracts expired. Both unions made growing concerns about the use of AI one of their top priorities. Unions have also embraced self-regulation. In recent years, it has become a standard practice for major unions to create AI commissions and institutes as well as establish and publish their own AI principles or guidelines.¹⁸18. See generally Amanda Ballantyne et al., Crafting an Innovation Ecosystem That Works for Working People, 34 New Eng. J. Pub. Pol’y 1 (2022).

This Article proceeds as follows: Part I examines the various ways the federal government has prioritized combatting AI if it is used for anti-union purposes or to displace workers, including executive orders and actions, interagency agreements, and specific agency actions. Next, Part II examines the legislative proposals that are being considered and the role unions have played in advancing these efforts at both the federal and state levels. Part III then turns to the new collective bargaining strategies and self-regulation efforts that unions are pursuing in response to the risks associated with AI. Finally, this Article concludes by providing some recommendations to help protect workers, employees, unions, and others without stifling innovation.

I. Federal Responses

A. Executive Orders and Actions

The Biden administration has used executive orders as a central part of its approach to employers using AI and emerging technologies for anti-union purposes. Most notably, on October 30, 2023, President Biden issued an “Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence” to address the growing concerns surrounding the use of AI.¹⁹19. Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, Exec. Order No. 14,110, 88 Fed. Reg. 75,191 (Nov. 1, 2023). The document directly addresses the risks of employers using AI for anti-union purposes. The fact sheet that accompanies the executive order identifies the risks involving workplace surveillance, bias, and job displacement, and it then explained that mitigating these risks must involve supporting “workers’ ability to bargain collectively.”²⁰20. Fact Sheet: President Biden Issues Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence, White House (Oct. 30, 2023), https://www.whitehouse.gov/briefing-room/statements-releases/2023/10/30/fact-sheet-president-biden-issues-executive-order-on-safe-secure-and-trustworthy-artificial-intelligence (https://perma.cc/6B4N-LXEC). Furthermore, one of the eight “guiding principles and priorities” of the executive order is explicitly pro-union: “The responsible development and use of AI require a commitment to supporting American workers. As AI creates new jobs and industries, all workers need a seat at the table, including through collective bargaining, to ensure that they benefit from these opportunities.”²¹21. Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, 88 Fed. Reg. at 75,192. Pro-union commentators praised the executive order for its efforts to signal that the government was promoting workers and unions, which, in their view, was more important than the actual content of the document.²²22. See Maddie Chang, Tech@Work — November 17, 2023, OnLabor (Nov. 17, 2023), https://onlabor.org/techwork-november-17-2023 (https://perma.cc/9XPE-4NSS) (“Perhaps more notable than the actual content of the sections pertaining to work is the prominence of work and workers’ interests in a policy document that might otherwise focus only on the defense related, economic, and civil rights elements of AI.”) (emphasis in original).

The executive order also requires federal agencies to take certain actions designed to address the risks of AI in the workplace. The executive order requires the U.S. Department of Labor (DOL) to issue a report on the impact of AI on job displacement. The order also requires DOL to “develop and publish principles and best practices for employers that could be used to mitigate AI’s potential harms to employees’ well-being and maximize its potential benefits.”²³23. Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, 88 Fed. Reg. at 75,210. Notably, to develop any standards or guidance related to AI, the order directs DOL to consult with “other agencies and with outside entities, including labor unions and workers, as the Secretary of Labor deems appropriate.”²⁴24. Id. Tellingly absent from this list is a direction that DOL consult with anyone from the employer community.

Specialized offices within the White House have also played a key role. The real or perceived proximity to the president provides these offices with significant influence and equips them with an inventory of both formal and informal tools of persuasion. In 2022, the White House’s Office of Science and Technology Policy (OSTP) issued its “Blueprint for an AI Bill of Rights.” The blueprint addresses contexts in which automation could lead to bias and discrimination, including at the workplace.²⁵25. White House Off. Sci. Tech. Pol’y, Blueprint for an AI Bill of Rights: Making Automated Systems Work for the American People 3 (2022), https://‌www.whitehouse.gov/wp-content/uploads/2022/10/Blueprint-for-an-AI-Bill-of-Rights.pdf (https://perma.cc/W73P-F6BH). It cites instances where employers had reportedly used “surveillance software to track employee discussions about union activity and use the resulting data to surveil individual employees and surreptitiously intervene in discussions.”²⁶26. Id. at 32. Critics of the “AI Bill of Rights” blueprint argue that it would likely necessitate the government to take aggressive steps to regulate AI to ensure adequate enforcement and thus hamper innovation and lead to increased regulatory adventurism.²⁷27. Sonderling et al., supra note 3, at 41. See also Adam Thierer & Neil Chilson, Overregulating AI Will Disrupt Markets and Discourage Competition, Bloomberg L. (Nov. 17, 2023, 9:00 AM CT), https://news.bloomberglaw.com/us-law-week/overregulating-ai-will-disrupt-markets-and-discourage-competition (https://perma.cc/FRR9-PNUT).

To further implement and advance the Biden administration’s approach to addressing unions’ concerns with AI, the White House has also turned to requests for information and listening sessions. In 2023, OSTP issued a request for information for public comments on the impacts of automated surveillance and management technologies on workers.²⁸28. OSTP, Request for Information; Automated Worker Surveillance and Management, 88 Fed. Reg. 27,932 (May 3, 2023). The request alleges, without any citation, that “[e]merging research suggests that certain applications of these systems may undermine . . . workers’ ability to organize and work collectively with their coworkers to improve working conditions, including through labor unions.”²⁹29. Id. at 27,933.

The White House has also used meetings and listening sessions to advance its approach to addressing unions’ concerns with AI. For instance, the White House hosted a listening session in June of 2023 with several high-profile union leaders to discuss the impact of AI on workers, job quality, and civil rights.³⁰30. Readout of White House Listening Session with Union Leaders on Advancing Responsible Artificial Intelligence Innovation, White House (July 3, 2023), https://www.whitehouse.gov/‌briefing-room/statements-releases/2023/07/03/readout-of-white-house-listening-session-with-union-leaders-on-advancing-responsible-artificial-intelligence-innovation (https://perma.cc/X9UC-5PN9). The listening session included officials from the White House National Economic Council, OSTP, and Office of the Vice President.³¹31. Id. Overall, the White House’s actions—including the Blueprint for an “AI Bill of Rights,” requests for information, and listening sessions—demonstrate an attempt to seize the narrative and shape the national discussion about AI’s impacts in order to place the focus on protecting jobs, workers, and unions.

B. Memoranda of Understanding

The federal government has also relied on interagency agreements, known as Memoranda of Understanding (MOUs). MOUs are generally unenforceable, non-binding agreements signed between various agencies that clarify agencies’ respective jurisdictions, assign regulatory tasks, and establish ground rules for information-sharing, investigation, training, enforcement, and other informal arrangements.³²32. See Kelley, All Along the New Watchtower, supra note 1, at 220-21. MOUs function as the network of contracts that aim to bring together interagency coordination within the administrative state and help streamline the process for investigating and penalizing businesses for a wide range of employer practices. MOUs impact states and local jurisdictions as well. Notably, DOL’s Wage and Hour Division (WHD) stresses that the agency uses MOUs to “build and maintain[] strong relationships with state and federal agencies to foster communication.”³³33. State Enforcement and Outreach Coordination, Wage & Hour Div. https://‌www.dol.gov/agencies/whd/about/state-coordination (https://perma.cc/ME84-NWMN) (last visited June 7, 2024). WHD administers and enforces the Fair Labor Standards Act, the Family and Medical Leave Act, and wage laws affecting government contractors. These MOUs are not limited to comparable state-level agencies. For example, WHD has MOUs with state attorney generals’ offices, counties, and district attorneys’ offices, and even non-governmental entities. Additionally, MOUs provide the federal government’s labor and employment agencies with a global reach because these agencies regularly enter into MOUs with foreign embassies and consulates. In 2022 alone, the U.S. Equal Opportunity Commission (EEOC), the Occupational Safety and Health Administration, WHD, the International Labor Affairs Bureau, and the NLRB signed MOUs with El Salvador, Guatemala, and Honduras.³⁴34. EEOC, DOL and Other Federal Agencies Sign Memorandum of Understanding with El Salvador, Guatemala and Honduras, U.S. Equal. Emp. Opportunity Comm’n (May 10, 2022), https://www.eeoc.gov/newsroom/eeoc-dol-and-other-federal-agencies-sign-memorandum-understanding-el-salvador-guatemala (https://perma.cc/8JRD-F5WY).

MOUs have been a critical tool for the federal government to address AI being used for anti-union purposes. In 2022, the NLRB and the Federal Trade Commission (FTC) signed an MOU which included a focus on “the impact of algorithmic decision-making on workers.”³⁵35. Lina M. Khan & Jennifer A. Abruzzo, Memorandum of Understanding Between the Federal Trade Commission (FTC) and the National Labor Relations Board (NLRB) Regarding Information Sharing, Cross-Agency Training, and Outreach in Areas of Common Regulatory Interest 1 (July 19, 2022), https://www.nlrb.gov/sites/default/files/attachments/pages/node-7857/ftcnlrb-mou-71922.pdf (https://perma.cc/MAA9-ZAHB). A year later, the NLRB and Consumer Financial Protection Bureau (CFPB) entered into an MOU stressing the need for the two agencies to address alleged harm underlying the gig economy. In a press release accompanying the MOU, the NLRB general counsel emphasized her belief that employers’ practices and use of AI-driven tools could chill workers from exercising their labor rights.³⁶36. CFPB and NLRB Announce Information Sharing Agreement to Protect American Consumers and Workers from Illegal Practices, CFPB (Mar. 7, 2023), https://www.consumerfinance.gov/about-us/newsroom/cfpb-and-nlrb-announce-information-sharing-agreement-to-protect-american-consumers-and-workers-from-illegal-practices/ (https://perma.cc/8JS7-P9MM). The general counsel stressed that the NLRB was “excited to work with the CFPB to strengthen our whole-of-government approach and ensure that employers obey the law and workers are able to fully and freely exercise their rights without interference or adverse consequences.”³⁷37. Id.

The NLRB has been the most active agency in pursuing MOUs to address anti-union AI risks. Most notably, in October 2022, the NLRB general counsel issued a memorandum emphasizing that the agency will use MOUs with several other federal agencies, including DOL, FTC, and the U.S. Department of Justice (DOJ), to facilitate coordinated enforcement against employers for their use of monitoring technologies.³⁸38. See Kelley, All Along the New Watchtower, supra note 1, at 220. The memorandum was released shortly after the White House released its “AI Bill of Rights” blueprint which briefly addressed the possibility of employers using this technology for anti-union purposes. This timing signaled that this was a White House priority and also underlined the close coordination between the White House and federal agencies.

The government’s strong focus on using MOUs raises several concerns. Connecting separate agencies that Congress established with responsibility for enforcing different laws outside of their specialized area generates noteworthy problems. For example, giving an agency a role in enforcing a law that is outside the scope of its specialty triggers confidentiality concerns.³⁹39. See id. at 226-27. One illustration of why this is problematic is a recent MOU that WHD entered with the EEOC in 2023. The MOU specifically contemplates that the EEOC may share certain employer information with WHD. Notably, Title VII of the Civil Rights Act of 1964, the Americans with Disabilities Act, and the Genetic Information Nondiscrimination Act prohibit the EEOC from disclosing certain data and information to the public, but the MOU does not bind WHD in the same way. Instead, the WHD agrees to “observe” the confidentiality requirements and the MOU provides an exception “in cases where WHD receives the same information from a source independent of the EEOC.” Memorandum of Understanding Between the U.S. Department of Labor, Wage and Hour Division and the U.S. Equal Employment Opportunity Commission, U.S. Equal. Emp. Opportunity Comm’n (Sept. 13, 2023), https://www.eeoc.gov/memorandum-understanding-between-us-department-labor-wage-and-hour-division-and-us-equal-employment (https://perma.cc/ML53-SX7U). When data is shared or complaints are referred between agencies, the receiving agency does not have the same familiarity with the confidentiality provisions. Critics of the government’s focus on interagency agreements contend that the agreements are self-serving for the government.⁴⁰40. For instance, these MOUs trigger concerns about what agencies are doing with information that they otherwise would not have a legitimate basis to obtain, or that otherwise seems totally irrelevant to the specific agency’s mission and is beyond the scope of the agency’s delegated authority. Critics argue that agencies should instead prioritize providing compliance assistance to the public so the regulated community can better understand how to comply with the law.⁴¹41. See, e.g., Kelley, All Along the New Watchtower, supra note 1, at 227.

C. Specific Agency Actions

The federal government has directed federal agencies to carry out specific acts to address the use of AI being used by employers for anti-union purposes. In many ways, the 2023 AI executive order established the mission and framework, the MOUs formed the necessary communication channels, and the agencies are implementing the tactical components of the larger AI strategy. Unsurprisingly, the NLRB is the most active agency attempting to address AI allegedly being used by employers for anti-union purposes. On October 31, 2022, the general counsel of the NLRB issued a memorandum that outlined a prosecutorial initiative aimed at employers that utilize technology to monitor and manage employees in the workplace.⁴²42. Memorandum GC 23-02, supra note 15. The GC’s Memo warns of “the potential for omnipresent surveillance and other algorithmic management tools to interfere with the exercise of Section 7 rights [to form or join unions] by significantly impairing or negating employees’ ability to engage in protected activity and keep that activity confidential from their employer.”⁴³43. Id. at 1. The GC’s Memo proposed a burden-shifting framework whereby an employer will be found to have presumptively violated the NLRA where its “surveillance and management practices, viewed as a whole, would tend to interfere with or prevent a reasonable employee from engaging in activity protected by the Act.”⁴⁴44. Id. at 8.

However, the proposed framework outlined in the GC’s Memo suffers from several flaws that undermine the approach that the general counsel proposes. First, the GC’s Memo does not distinguish lawful from unlawful monitoring and leaves critical terms undefined, thereby proposing a standard that is almost impossible to meet.⁴⁵45. See Kelley, All Along the New Watchtower, supra note 1, at 222-23. Second, the GC’s Memo fails to account for the wide diversity of AI tools and the many legitimate business purposes for employee monitoring, including detecting and mitigating cybersecurity threats, ensuring compliance with workplace guidelines, preventing discrimination, harassment, and workplace violence, and enhancing workplace health and safety.⁴⁶46. Id. at 223. For example, AI-powered agricultural equipment has been shown to improve safety by reducing how many workers are needed for labor-intensive tasks during hot weather and by removing operators from hazardous tasks such as moving a pesticide sprayer.⁴⁷47. See Bruce Rolfsen, AI-Based Farm Equipment May Increase Worker Safety but Cost Jobs, Bloomberg L. (June 27, 2023, 4:30 AM CT), https://news.bloomberglaw.com/safety/ai-based-farm-equipment-may-increase-worker-safety-but-cost-jobs (https://perma.cc/3QF8-E6YP). Third, the position outlined in the GC’s Memo also fails to recognize that many AI practices at issue are driven by compliance with several employment laws and regulations, particularly in the areas of anti-discrimination, anti-harassment, and occupational health and safety. For instance, AI-driven tools can be used to mitigate work-related violence and harassment risks by detecting patterns and identifying or predicting risks to find the best way to minimize such risks.⁴⁸48. See Leora Eisenstadt, #MeTooBots and the AI Workplace, 24 U. Pa. J. Bus. L. 350, 352 (2022) (describing AI tools known as #MeTooBots that monitor and flag communications between colleagues to address the problem of sexual harassment in the workplace). Fourth, because many employers have increased their use of technological tools to effectively manage their increasingly off-site workforces, the GC’s Memo will also impair remote work and therefore hurt employee morale, retention, and productivity.⁴⁹49. See Kelley, All Along the New Watchtower, supra note 1, at 226. Ultimately, the regulatory framework outlined in the GC’s Memo, which was issued without notice-and-comment, will likely harm the workers the general counsel purportedly seeks to protect.

Other federal agencies have sought to address the use of AI for anti-union purposes. In March of 2022, the CFPB released a blog post detailing a recent meeting in which CFPB invited worker organizations and labor unions representing workers to share their members’ experiences and challenges.⁵⁰50. See Emma Oppenheim, Shining a Spotlight on Workers’ Financial Experiences, CFPB (Mar. 9, 2022), https://www.consumerfinance.gov/about-us/blog/shining-a-spotlight-on-workers-financial-experiences (https://perma.cc/JF99-Y423). CFPB’s blog post stated that organizations reported concerns about the rise of new surveillance technology tools that are enabling the collection of unprecedented amounts of information about workers. The blog post further stated that participants at the meeting raised concerns regarding how the information is compiled and used for decision-making that may impact workers’ financial well-being in the long run.⁵¹51. Id.

II. Legislative Efforts

Federal legislative proposals have sought to address the accelerated deployment of AI and other emerging technologies by employers. Several union-supported proposals are being considered. One of the most far-reaching proposals is the Stop Spying Bosses Act which would require disclosures and prohibit employers from engaging in surveillance of workers.⁵²52. Stop Spying Bosses Act, S. 262, 118th Cong. (2023). More specifically, the bill would expressly prohibit employers from surveilling workers engaged in union or protected labor activity, and it would require companies to disclose key aspects of their surveillance and data collection practices. The bill has been endorsed by several leading unions, including the AFL-CIO, the Communications Workers of America, and Service Employees International Union.⁵³53. Casey, Booker, Schatz Introduce Bill to Protect Workers from Invasive, Exploitative Surveillance Technologies, Bob Casey (Feb. 2, 2023), https://www.casey.senate.gov/news/releases/casey-booker-schatz-introduce-bill-to-protect-workers-from-invasive-exploitative-surveillance-technologies (https://perma.cc/ZA6F-KCSL). Although no legislation has been passed yet at the federal level, these attempts to implement AI laws reveal Congress’s desire to regulate this technology in the future.

Like the legislation being considered at the federal level, union-friendly legislators and unions themselves are promoting state-level strategies as states and local jurisdictions have increasingly sought to fill the federal void by enacting their own AI legislation.⁵⁴54. Williams, supra note 16 (noting that some union leaders argue that while federal efforts can “set a tone and a framework for the states,” state level action should go beyond). For example, the International Brotherhood of Teamsters has advocated for a bill in California which would require a human operator in autonomous vehicles.⁵⁵55. Assemb. B. 316, 2023 Cal. State Assemb. (Cal. 2023). Although the California bill was vetoed by the governor, attempts to implement these types of laws reveal the unions’ strong desire to regulate AI technology to benefit their members.

Most of the state proposals focus on worker displacement. For instance, New Jersey is considering a proposal that would require the state’s Department of Labor and Workforce Development to track and maintain information on job loss due to automation.⁵⁶56. Assemb. B. 5150, 2023 Gen. Assemb. (N.J. 2023). The bill would also require the state’s Department of Labor and Workforce Development to identify industries and positions that are vulnerable to job loss and to provide resources for individuals at risk for job loss and who may be qualified for alternative positions with or without training, education, or other experience.⁵⁷57. Id.

To date, some states have successfully passed laws addressing the impact of emerging technologies on the workforce. For example, in 2023, Colorado passed a law requiring the state’s Office of Future of Work to contract with a third party to study workforce transitions in the state’s economy, including the skill transferability of workers in occupations facing the most disruption due to automation.⁵⁸58. Colo. Rev. Stat. Ann. § 8-83-902 (West 2023). The Colorado law is focused on workers in the oil and gas industry, but the law affects other workers as well. The law also requires the contractor to provide recommendations for programs and policies that may be required to prepare employers and workers for any transitions.

Legislators and unions have also focused on specific industries, such as the entertainment industry. As discussed above, the entertainment industry has been a particular focus of unions because of heightened concerns that computerized simulations of actors could replace the actors themselves and writers would be relegated to refining AI-generated scripts, especially in the age of ChatGPT. In New York, SAG-AFTRA is supporting a bill that would ban a state tax credit for any TV or movie production that uses AI to displace workers.⁵⁹59. S.B. 7422, 2023 N.Y. State S. (N.Y. 2023). See also Williams, supra note 16. In addition, SAG-AFTRA and the California Labor Federation, which represents 1,300 unions and over two million union members, are supporting a bill in California that would bar employers from negotiating to use a worker’s voice or likeness to create a digital replica or train AI unless the worker is represented by an attorney or union representative.⁶⁰60. Assemb. B. 2602, 2024 Cal. State Assemb. (Cal. 2024); Assemblymember Kalra, SAG-AFTRA Introduce New Bill on Generative AI in Performer Contracts, Assemblymember Ash Kalra (Feb. 14, 2024), https://a25.asmdc.org/press-releases/20240214-assemblymember-kalra-sag-aftra-introduce-new-bill-generative-ai-performer (https://perma.cc/7FXL-SDFV).

III. Collective Bargaining and Private Initiatives

In the absence of comprehensive regulation, AI’s increasing ubiquity and expanding commercial potential has led some unions to turn to collective bargaining to address the risks associated with AI. Collective bargaining refers to the process where a union negotiates with employers on behalf of its members to establish terms of employment, such as wages, hours, benefits, and working conditions. This negotiation aims to reach a collective agreement that purports to balance the interests of both the employees—represented by the union—and the employer. Union leaders argue that collective bargaining should be used to determine whether new technologies can (or should) be introduced into workplaces, minimize any disruptions on workers, and enhance the effective adoption of workplace technologies.⁶¹61. See Ballantyne et al., supra note 18, at 1, 2. Some unions have attempted to add entirely new protections into their labor contracts while other unions have focused on updating existing language to account for AI and other emerging technologies. Two notable instances occurred within entertainment industry when the Writers Guild of America (WGA) and SAG-AFTRA went on prolonged strikes following the expiration of their labor contracts. Both unions made concerns about the use of AI part of their key priorities.

The Hollywood strikes and the collective bargaining negotiations were noteworthy for attempting to enshrine into their labor contracts entirely new protections to address AI. In October 2023, WGA announced that it reached an agreement on the terms of a new contract with the major Hollywood production studios. That contract contains what many consider to be the first major union-management agreement regulating the role of AI across a particular industry.⁶²62. See Matt Scherer, New WGA Labor Agreement Gives Hollywood Writers Important Protections in the Era of AI, Ctr. Democracy & Tech. (Oct. 17, 2023), https://cdt.org/insights/new-wga-labor-agreement-gives-hollywood-writers-important-protections-in-the-era-of-ai (https://perma.cc/F99V-4565). The WGA contract included assurances that AI cannot be used to undermine a writer’s credit as well as requirements for Hollywood studios to disclose if any material provided to a writer was produced by AI, while allowing writers to individually use AI tools if they choose. Under the agreement, even if a studio or an individual screenwriter uses AI tools to generate an idea for a new film, credit for coming up with that story cannot go to the AI system or whoever developed or deployed it. The agreement also allows writers to use generative AI, but they must follow studio policies if they do so. Commentators argue that the WGA agreement could serve as a model for future labor agreements involving generative AI and automation more generally and “shows that collective bargaining can be an effective means for workers to advance their interests when automation threatens their livelihoods.”⁶³63. Id.

In November 2023, SAG-AFTRA reached a tentative agreement with Hollywood studios and the use of AI was divided into three new categories.⁶⁴64. See generally SAG-AFTRA, TV/Theatrical Contracts 2023: Summary of Tentative Agreement (Nov. 16, 2023), https://www.sagaftra.org/files/sa_documents/TV-Theatrical_23_Summary_Agreement_Final.pdf (https://perma.cc/2LQM-RZB5). The first two categories include “employment related” and “independently created” digital replicas. Employment-related replicas are those based on an actor’s physical performance in a movie or TV show. By contrast, independently created replicas are those based on an actor’s general likeness, voice, or features, but not a specific performance. For both kinds, a studio must notify the actor, get the actor’s consent, and pay the actor as if the actor were personally performing the replica’s role. A third category, “digital synthetics,” includes wholly digital actors not based on any natural person. A studio can use a digital synthetic without consent. It must, however, notify the union and give the union a chance to bargain. Further, the contract requires studios to meet twice per year with the union to re-discuss the evolving uses of generative AI. The contract gives the same notice and compensation rights to principal and background actors. It also gives the union a right to meet twice a year with each studio. At these meetings, the studio will update the union about its plans to use AI.⁶⁵65. Id.

The SAG-AFTRA contract also has some important limitations. For one, it limits actors’ remedies. It gives actors no rights to retroactively veto or block projects. Instead, the most an actor can do is recover unpaid compensation. The contract also gives the studios leeway to modify an actor’s performance with AI in post-production. For example, a studio can modify the actor’s lip movements to match a foreign language. The studio does not need the actor’s consent to do that. Nor does it need to pay the actor anything extra. Some have argued that the provisions of the SAG-AFTRA contract leave studios free to replace actors with digital creations. These members would have preferred a complete ban on generative AI. In response, the union released a frequently asked questions document addressing AI-related concerns.⁶⁶66. SAG-AFTRA, FAQs on AI 1 (2023) https://www.sagaftra.org/files/sa_documents/AIFAQs.pdf (https://perma.cc/YJ39-XA6T) (noting that “(a) ban on AI, while it sounds good in theory, could have significant unintended consequences, particularly as we cannot control companies with whom we don’t have a bargaining relationship”). The document, in effect, argues that the union got everything it could while also suggesting that the union may seek tighter restrictions later.⁶⁷67. Id. at 1, 6. Regarding the issue of banning companies from training AI on SAG-AFTRA members’ work: “This is a topic we will continue to monitor and, depending on the outcome of the litigation and studies, we can revisit this topic if the courts and copyright office find training requires consent.” Id. at 1.

Meanwhile, other unions have focused on strengthening and updating existing labor contract language to account for evolving technology. For instance, the Culinary Union—the union representing Las Vegas service and hospitality workers—reached an agreement with three large casinos that requires the casinos to notify the union before introducing or modifying new technology (including AI and robotics), and requires the casinos to bargain over any technology used for tracking workers or monitoring their performance.⁶⁸68. Ian Kullgren, Las Vegas Union Scores AI, Daily Cleaning Wins in Caesars Pact, Bloomberg L. (Nov. 8, 2023, 6:41 PM CT), https://news.bloomberglaw.com/daily-labor-report/las-vegas-union-scores-ai-daily-cleaning-wins-in-caesars-pact# (https://perma.cc/7LJ3-J3JZ). The contract also offers displaced workers protections such as new pay, benefits, and recall rights.

Unions have also created their own private initiatives to address the challenges associated with AI and emerging technologies at the workplace. For instance, the AFL-CIO, the largest federation of labor unions in the United States, both created a Commission on the Future of Work and Unions and launched a Technology Institute in 2021 that seeks to bring together unions, universities, think tanks, and other organizations.⁶⁹69. See Ballantyne et al., supra note 18, at 3. Similarly, in 2023, the International Alliance of Theatrical Stage Employees (IATSE), a labor union representing over 168,000 technicians, artisans and craftspersons in the entertainment industry, created a Commission on Artificial Intelligence.⁷⁰70. IATSE Releases Core Principles for Applications of Artificial Intelligence and Machine Learning Technology, IATSE (July 5, 2023), https://iatse.net/iatse-releases-core-principles-for-applications-of-artificial-intelligence-and-machine-learning-technology (https://perma.cc/A9DL-65V7).

In recent years, it has become a standard practice for major unions to institute and publish their own AI principles or guidelines. For instance, in 2023, the IATSE released its “Core Principles for the Application of Artificial Intelligence (AI) and Machine Learning (ML) technologies in the entertainment industry.”⁷¹71. Id. Those principles include research, collaboration with partners and stakeholders, education, political and legislative advocacy, and collective bargaining. The IATSE’s principles do not disparage AI but instead focus on ensuring that the “fruits of increased productivity through AI are shared equitably among all stakeholders.”⁷²72. IATSE, Core Principles for Applications of Artificial Intelligence and Machine Learning Technology 1 (July 5, 2023) The principles also encourage self-regulation through collective bargaining, stressing that “collective bargaining is the primary way to ensure workers do not have to wait for government regulation through legislation, which could take years or may never come at all.”⁷³73. Id. at 2.

Another recent development has been unions forming partnerships with major tech companies to address AI. Most notably, in December 2023, the AFL-CIO and Microsoft announced a new partnership to discuss the use of AI in the workplace.⁷⁴74. AFL-CIO and Microsoft Announce New Tech-Labor Partnership on AI and the Future of the Workforce, AFL-CIO (Dec. 11, 2023), https://aflcio.org/press/releases/afl-cio-and-microsoft-announce-new-tech-labor-partnership-ai-and-future-workforce (https://perma.cc/4PS2-SL4G). The partnership has three goals: (1) sharing information with labor leaders and workers on AI technology trends; (2) incorporating worker perspectives and expertise in the development of AI technology; and (3) helping shape public policy that supports the technology skills and needs of frontline workers.

IV. Recommendations

Because AI regulatory efforts are still in an early stage, this Part of the Article outlines some key recommendations that can help protect workers, employees, unions, and others without stifling innovation. First, any federal agency regulatory efforts should be done through notice-and-comment rulemaking. Notice-and-comment is particularly important with newer workplace technologies and AI because public comments can help improve the guidance by providing outside parties the opportunity to provide meaningful feedback, including pivotal responses from industry experts.⁷⁵75. Kelley, All Along the New Watchtower, supra note 1, at 233-34. Ultimately, clear, comprehensive, and reasonable guidance that is enforced predictably and consistently will help encourage employers, technology vendors, and others to proactively prevent the negative effects of these technologies. Moreover, such guidance reduces uncertainty and protects workers, employees, applicants, unions, and others without stifling innovation.

Agencies seeking to regulate AI without notice-and-comment are operating in an echo chamber resulting in regulations that not only lack diverse points of view but are also devoid of input from subject matter experts. The regulatory framework proposed in the NLRB general counsel’s 2022 AI memorandum is one example of the consequences of failing to receive any stakeholder input. It failed to recognize that many AI practices at issue are driven by compliance with several employment laws and regulations, particularly in the areas of anti-discrimination and occupational health and safety. The EEOC’s approach to AI is another illustration of a flawed approach to addressing AI. The EEOC launched an initiative in 2021 to ensure that AI and other emerging tools used in hiring and other employment decisions comply with the federal civil rights laws that the agency enforces. However, the agency has failed to issue any AI guidance documents that have been subject to notice-and-comment.⁷⁶76. Sonderling et al., supra note 3, at 42. See also Keith E. Sonderling & Bradford J. Kelley, Filling the Void: Artificial Intelligence and Private Initiatives, 24 N.C. J.L. & Tech. 153, 161 (2023). In addition, the EEOC failed to include any vendors involved with the development of AI solutions in the single non-technical public hearing it has conducted. As a result, the employers who often buy AI tools from these vendors have little insight into how the technology works despite facing the greatest potential liability in the event of a lawsuit.⁷⁷77. Id. at 161 n. 28. At bottom, the submission of public comments fundamentally helps improve the regulatory process by providing outside parties, including trade associations, employers, unions, civil rights groups, and others, the opportunity to provide meaningful feedback.

Second, policymakers should collaborate with the private initiatives that have been created by companies, unions, and other organizations to facilitate shared goals. Businesses, unions, and non-governmental organizations must be at the vanguard of our national discussions on AI to ensure that it is developed and deployed responsibly and consistent with our shared values. To do so, these groups should collaborate with the government to ensure the responsible use of AI in the workplace. These private initiatives can facilitate dialogue among unions and employers to ensure that workers can fulfill existing roles and pivot as necessary as new roles emerge. Indeed, traditional regulatory frameworks are oftentimes draconian and slow to adapt to the complicated and rapidly evolving technologies, whereas private initiatives are flexible and can respond quickly to the breathtaking pace of AI development. The Biden administration has recognized the importance of these self-regulatory mechanisms while developing the administration’s approach to addressing AI for anti-union purposes. Notably, in 2023, the White House announced that it had secured voluntary commitments from fifteen of the leading AI companies to control the risks posed by AI.⁷⁸78. Fact Sheet: Biden-⁠Harris Administration Secures Voluntary Commitments from Eight Additional Artificial Intelligence Companies to Manage the Risks Posed by AI, White House (Sept. 12, 2023), https://www.whitehouse.gov/briefing-room/statements-releases/‌2023/09/12/fact-sheet-biden-harris-administration-secures-voluntary-commitments-from-eight-additional-artificial-intelligence-companies-to-manage-the-risks-posed-by-ai (https://perma.cc/89W7-YMM8). During a White House listening session on advancing responsible AI innovation, several influential union leaders, including the leader of the AFL-CIO’s Technology Institute, “shared views on possible opportunities for AI to improve workers’ lives when unions and workers are at the table and jointly developing solutions with employers.”⁷⁹79. Readout of White House Listening Session with Union Leaders on Advancing Responsible Artificial Intelligence Innovation, supra note 30. At a minimum, a joint effort between private initiatives and public institutions is needed to create a more agile regulatory framework that is fully responsive to the accelerating pace of emerging technologies. Moreover, private initiatives can undoubtedly help build a culture of trust, transparency, and accountability in AI technologies. As a result, any AI regulatory efforts should include key guidance and workable directives developed in cooperation with private initiatives.

Third, policymakers must be aware that the growing patchwork of laws across the nation and conflicting agency requirements present compliance challenges for employers. Conflicting government compliance requirements for companies create a complex web of challenges that can negatively impact employers, employees, and unions, affecting everything from job security and workplace conditions to the unions’ ability to effectively represent and negotiate on behalf of their members. As such, policymakers must seriously consider the value of a national standard that simplifies regulatory compliance and preempts conflicting regulatory frameworks at the state and local levels. This is of particular importance for employers with multi-state or multi-national operations, who are concerned that a proliferation of state and local laws will set forth innumerable compliance regimes (which are rarely consistent and potentially conflicting).

Conclusion

Ultimately, the existence of the federal void demonstrates the fact that much of what the administration has done is focused on messaging and signaling, as opposed to actionable, concrete regulatory steps. In a way, the administration is shrewdly walking the line by being very vocal in its support for labor, while simultaneously not pushing for aggressive regulations that would risk the ire of businesses and employers.

Regulating AI will continue to be debated for the foreseeable future. Employers, employees, unions, policymakers, and others must carefully navigate the intersection of technological advancements and labor and employment laws to shape the future of AI regulation. Because AI is advancing rapidly, it will become increasingly difficult to address these issues if the can is kicked down the road. In the absence of comprehensive regulation, self-regulation based upon existing laws and innovative applications offers the most viable path forward to address the issues raised by AI now and in the future.

†  Shareholder, Littler Mendelson P.C. For helpful comments and advice on prior drafts, the Author thanks Tessa Gelbman, Mike Skidgel, Allan King, and Maureen Lavery.