The labor movement has a special responsibility to confront artificial intelligence’s imposition on workers: without unions, bosses have carte blanche to use AI to undercut workers at every level.
When Boston University graduate students went on strike in April, Stan Sclaroff, the university’s dean of arts and sciences, sent faculty an email with suggestions for keeping their classes on track. As Inside Higher Ed reported, the dean’s “creative” solutions included combining discussion sections, alternative assignments, and using “generative AI tools like ChatGPT.” Professors, the dean wrote, could use the technology to “give feedback or facilitate ‘discussion’ on readings or assignments.”
Sclaroff’s suggestion that artificial intelligence (AI) could handle struck work provoked an outcry from grad students and faculty, with Service Employees International Union (SEIU) Local 509, which represents the grad students, stating, “We are extremely disappointed by the university’s suggestion that the use of AI could even begin to substitute the hard work that graduate workers pour into mentoring students, facilitating discussions and teaching.” The union told Inside Higher Ed that they hoped the administration would reconsider the suggestion and instead “focus on properly compensating the people who do the work that is crucial in keeping the university running.” The university soon backtracked, claiming, “Neither Dean Sclaroff nor Boston University believe that AI can replace its graduate student teaching assistants, and the assertion that we plan to do so is patently false.”
Since ChatGPT launched in November 2022, educators are just one of many workers now confronted with emerging technology wielded by management as a threat. Across the United States, AI has occasioned all manner of anxieties. It’s not just AI’s environmental impacts, discriminatory biases, or privacy risks: according to one study from Ernst & Young, a consulting firm, some 75 percent of workers are concerned AI will make certain jobs obsolete, with 65 percent worried about AI replacing their job. An Axios-Morning consult poll found that while regulation of the new technology is not a top priority, some half of Americans worry that AI will replace either their job or those of their loved ones.
There’s reason for concern. Despite AI’s potential for liberating workers from tedious tasks and grunt work — proofreading, drafting standard-operating procedures, coding data, identifying images, and much more — in the hands of employers, the technology is already being used to replace workers and undermine their bargaining power. Tech companies often employ low-paid workers in the Global South to correct AI’s mistakes, raising the question of just how “artificial” the technology really is. Reliable numbers are hard to come by as employers seek to avoid outcry from replacing workers with AI, but according to data from outplacement firm Challenger, Gray, & Christmas, an estimated minimum of 4,000 US workers lost their jobs to AI last year. As AI pioneer Mustafa Suleyman recently put it, “Left completely to the market and to their own devices, [AI tools are] fundamentally labor-replacing.” Dean Sclaroff was merely making explicit a dynamic that already exists, and in a mind-boggling variety of industries. Everywhere one looks, workers are confronting AI not as a means of reducing the tedium of their work, but as a boon to managerial power.
Take the fast-food industry. In 2019, Nation’s Restaurant News was already calling the technology the “biggest tech trend” in quick-serve restaurants (QSRs), where experiments with AI are underway in both the front of the house (servers, bartenders, hosts) and back (busboys, cooks, dishwashers). Companies have been adapting conversational AI for drive-through windows for years (with occasionally disastrous results shared on social media), but as the publication noted, “the trend has now reached a fever pitch, and more and more limited-service chains are replacing humans with AI drive-thru workers.” The list of chains experimenting with the technology either for drive-through or online ordering include Wendy’s, IHOP, CKE Restaurants, White Castle, Del Taco, Panera, McDonald’s, and Sonic. Some restaurant owners are dreaming of automated servers for sit-down dining; while robot servers are a rarity in the United States, in China, Pizza Hut already has some one thousand of them.
Those are just customer-facing roles. In the back of the house, QSRs are toying with deploying AI for inventory management, food safety, and worker scheduling, while a few are testing robots that can prepare hamburgers and turn out tortilla chips. The tight labor market of recent years has led employers to search for alternatives to paying workers the higher compensation that they can now command, hoping AI might be the latest labor-saving technological investment that can boost long-run profits.
That process may already be unfolding at UPS. Earlier this year, the company announced that it plans to lay off 12,000, the largest layoffs in the company’s history. According to CEO Carol Tomé, the cuts, which will primarily affect management-level employees rather than the company’s unionized workers, were in part made possible by AI. She cited the example of the technology allowing salespeople to put together proposals without having to ask pricing experts for guidance as one way AI can reduce the logistics giant’s headcount.
This year, the meat-packing industry’s annual Meat Conference featured a panel titled “The Use and Application of Artificial Intelligence: How to Make this Advanced Technology Work for Your Business.” According to the organization, the presentation aimed to “provide industry stakeholders with insights on how data application can deliver operational improvements, like enhanced efficiencies and a more reliable supply chain, while also better predicting shopper habits and needs.”
As with any new technology, it’s difficult to cut through the hype around AI, such as OpenAI CEO Sam Altman’s claim that AI will become “the most powerful technology humanity has yet invented” or former Google chief business officer Mo Gawdat’s boast that “We’re creating God.” Tech companies have an interest in exaggerating AI’s capabilities. Silicon Valley executives have long insisted that autonomous vehicles would replace truck drivers, but they have hardly done so. Instead, truckers are being transformed into appendages of technology, forced to drive under the watchful eye of surveillance technologies that monitor and control them rather than render them obsolete.
AI is likewise being used to monitor workers, a dynamic dramatically on display in the growing use of technology by Aware, an AI firm specializing in analyzing employee text messages and emails. The company told CNBC that employers including Walmart, Delta, T-Mobile, Chevron, and Starbucks are now using its tools to monitor employees’ messages on Slack, Zoom, and other workplace-applications. Aware describes its product as a means of analyzing employee sentiment and toxicity; it isn’t hard to imagine how it might aid employers hoping to halt worker organizing. Rather than replacing them outright, AI’s biggest threat may be as an innovation in management technologies, with workers governed by decisions made by algorithms while simultaneously surveilled by other algorithms.
“A lot of this becomes thought crime,” Jutta Williams, cofounder of Humane Intelligence, told CNBC of AI employee surveillance technology. “This is treating people like inventory in a way I’ve not seen.”
In an AI-centric future, we are sure to see some jobs, both skilled and unskilled, replaced by automation: educators, fast-food workers, warehouse workers, software engineers, graphic designers, writers, and even lawyers are all on the chopping block. The technology will create new jobs as well: tending to the algorithms, producing them, fixing their mistakes. Microsoft CEO Brad Smith has said that the goal of AI is to “‘boost the productivity of workers, [reduce] the drudgery in jobs’, and translates those efficiency gains in higher standards of living,” but many jobs will be made much more onerous and less fulfilling, as workers are forced to coexist with AI, correcting its errors and reworking its output.
To understand the scope of AI’s potential impact, look at the numbers. Investment in AI is rapidly accelerating, and employers across industries (as well as financial analysts) are enamored with its possibilities. A 2022 study found that the amount of computing power used to train AI systems has been doubling every six months over the past decade.
Cognitive workers are a particular target, as generative AI systems are being trained to create computer code, write sentences and emails, summarize articles, brainstorm ideas, and translate languages. (Research by OpenAI, the company behind ChatGPT, states that language-learning models [LLMs] could affect 80 percent of the US workforce in some capacity — perhaps an overstatement, given the company’s interest in promoting its products, but also a sign of the extent of the company’s ambition.) An emerging consensus suggests that AI’s productivity effects will be significant, with changes already evident at call centers as the technology is deployed to answer common questions, route customers to agents, and generate phone-call transcripts.
AI is fueling change in American workplaces, and that change is coming fast. Government regulations, however, are lagging. President Joe Biden issued an executive order on AI in October 2023 which states that “as AI creates new jobs and industries, all workers need a seat at the table, including through collective bargaining.” Biden directed federal agencies to “develop principles and best practices to mitigate the harms and maximize the benefits of AI for workers by addressing job displacement; labor standards; workplace equity, health, and safety; and data collection,” a means of “providing guidance to prevent employers from undercompensating workers, evaluating job applications unfairly, or impinging on workers’ ability to organize.”
But as agencies develop such standards, AI’s rollout continues, and with union density at 10 percent overall and a ghastly 6 percent in the private sector in 2023, it is doing so while the overwhelming majority of workers don’t have the seat at the table of which Biden’s order speaks. The PRO Act, a labor-law reform package that would make union organizing far easier, remains stalled in Congress, and that means they aren’t likely to get one anytime soon. The result? Those workers who do have unions are left to battle AI’s anti-worker impacts on their own.
Unions Take on AI
The class war over AI entered the mainstream last year. When workers across the entertainment industry went on strike, constraining film studios’ use of the technology was a top priority. For the 11,500 striking screenwriters and television writers in the Writers Guild of America (WGA) and the 160,000 striking actors and performers in the Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA), the threat was imminent. In Hollywood, tales abound of producers wanting performers to scan their likeness or voice for possible future use. During the strike, actors told me of their fears that digital versions of themselves could be made to say and do things they never would have, with audiences unable to tell the difference. Writers worried they might become mere adjuncts to AI-written plots or scripts, hired to punch up a story produced by a technology that itself is trained on writers’ former work (a dynamic that itself raises legal questions, given authors’ copyright claims).
Indeed, the threat AI poses to performers is playing out dramatically in a dispute between actress Scarlett Johansson and OpenAI. The company is launching an automated voice assistant, and according to Johansson, Sam Altman, the company’s CEO, asked her last year if she would lend her voice to the tool (Johansson played a disembodied AI voice in Spike Jonze’s Her, which Altman has said is his favorite movie). She declined, but when the tool launched, it sounded remarkably similar to her anyway, a move that she said “shocked” and “angered” her.
“In a time when we are all grappling with deepfakes and the protection of our own likeness, our own work, our own identities, I believe these are questions that deserve absolute clarity,” Johansson said of the dispute. “I look forward to resolution in the form of transparency and the passage of appropriate legislation to help ensure that individual rights are protected.” Following Johansson’s statement, OpenAI announced it would remove the voice from the tool, though the company claims the voice is not an imitation of the star but rather comes from a different professional actress using her natural speaking voice.
It is precisely this theft of one’s self, the possibility of losing control over one’s voice or likeness, which motivated writers and actors to stay out on the picket lines. The contracts they won secured a preliminary goal: creating a framework of regulations and standards for its use, contractual mandates by which film studios and producers must abide.
After 148 days on strike, WGA members ratified a contract which stipulates that AI is not a person, and thus, its output cannot be considered “literary material.” A company can still furnish a writer with AI-generated material and ask them to work from it, but this cannot affect either the writer’s compensation or credits, a specification that reduces companies’ incentives to use AI in the first place. Per the contract, a company also cannot require a writer to use generative AI themselves, be it ChatGPT or any other program.
The SAG-AFTRA contract, ratified after 118 days on strike, offers more extensive AI rules than the writers’ agreement, reflecting the fact that the technology has already affected members. Studios are now required to obtain performers’ consent before a “digital replica” or their voice or likeness can be created or used; when it is used, they must compensate performers at a rate similar to what performers would receive if they were on set themselves.
Film-industry workers aren’t the only ones bargaining over AI. In November 2023, Las Vegas hospitality workers, members of the Culinary union, secured expanded technology rights, including retraining for those with jobs altered or replaced by AI. Journalists are yet another front line: while the New York Times is suing OpenAI and Microsoft for copyright infringement, claiming that the newspaper’s articles were used to train the technology, other publishers, including Vox Media, News Corp (publisher of the Wall Street Journal), the Atlantic, and the Financial Times have inked deals with OpenAI, and a range of publications are experimenting with AI-generated articles. Accordingly, both the WGA and the NewsGuild, the country’s two biggest unions for journalists, have demanded protections from AI in writing. The Financial Times Guild, a WGA-East shop, was the first to win the right to bargain over the effects of newly introduced technology like AI. According to David Isenberg, who covers the Securities and Exchange Commission for FT Specialist, the company wanted to limit workers’ rights to merely discussing changes rather than having a mandate to bargain over them, but members insisted on the legal obligations of a right to bargain and their first contract, ratified last summer, includes that language.
The Communications Workers of America (CWA), the NewsGuild’s parent union, recently established principles for AI in the workplace: accountability, proactive bargaining, and early and meaningful worker input. As CWA not only represents journalists, but workers in call centers, telecommunications, and tech — the Alphabet Workers Union-CWA includes members who work on AI systems at Google and other Alphabet, Inc. companies — the union is well-positioned to establish standards worth adopting across the labor movement.
“AI has the potential to build prosperity and unleash human creativity, but only if it works for working people,” CWA president Claude Cummings Jr said in a statement announcing the AI principles. “We are taking a member-first approach and demanding that working Americans have a voice, guaranteed by their union contract, in how AI shapes the future of work.” The union’s organizing at Microsoft itself, where it represents hundreds of quality-assurance workers at subsidiary ZeniMax Media, a video game holding company, offers one example of what these statements look like in practice.
The International Brotherhood of Teamsters has made similar statements, with the New York State Conference of Teamsters testifying that “any type of artificial intelligence must be regulated to protect the public by requiring safe operation and meaningfully address what many, including the Teamsters, view as a potentially catastrophic impact on the American workforce.” The union singled out autonomous vehicles as a threat to their members, though the recently ratified UPS contract, despite securing major wins, does not halt the company from acquiring a fleet of autonomous vehicles.
“Bring workers in early in the process,” American Federation of Labor and Congress of Industrial Organizations (AFL-CIO) president Liz Shuler said in December 2023. “Use our expertise to brainstorm, develop and implement new ideas.” The labor federation has been advocating for a seat at the table in AI’s rollout and is also pushing for worker input in the production of the technology itself, what Shuler calls a “co-creation process.” Those aims are reflected in a recent partnership between the AFL-CIO and Microsoft, which includes a mechanism for labor to share insights and concerns with those developing the technology at the company, as well as a plan for the parties to formulate joint policy recommendations (the agreement also mandates Microsoft remain neutral if any of its workers organize).
There are other experiments in worker input too: for instance, the labor federation’s Technology Institute has had union bus drivers offer their perspectives on self-driving buses, resulting in a white paper by Carnegie Mellon University professors who realized that the technology would complicate operators’ jobs. The white paper also noted that “self-driving technology is no substitute for the more human part of what bus operators do in attending to passenger needs (say, assisting an elderly person) and looking out for passenger safety.”
A few years ago, Shuler and a delegation of UNITE HERE members visited Carnegie Mellon University, a technology powerhouse, to better understand AI. A faculty member demonstrated a “cutting-edge cutting board that was able to measure how fast someone sliced vegetables, as well as the quality of their motion.” As Shuler told the publication, “It was easy to see we were basically one step away from having the technology cut the vegetables.”
Stop AI or Tame It?
When SAG-AFTRA leadership announced the details of the union’s tentative agreement with the studios last year, some actors (and even a few of the union’s board members) weren’t satisfied with the protections against AI and urged members to vote against ratification. The contract allows for exceptions to its AI compensation and consent requirements, such as when an actor’s digital replica is used for “comment, criticism, scholarship, satire or parody,” an ambiguous and potentially broad carveout. Others objected to countenancing the use of AI at all, arguing that the technology is an unambiguous threat to current and future members, and the union’s role should be to oppose it, full stop.
It was an early instance of a debate that will continue to play out as employers’ adoption of AI continues: Should unions try to stop AI, or merely secure a say in its use and maybe development? Can unions’ efforts to keep the technology at bay neutralize its anti-worker effects, assuaging the fear of job loss that keeps so many workers up at night? Or are the SAG-AFTRA critics right, and if there ever were a chance to produce AI that benefited society at large rather than a minority of employers and workers, it has long gone?
This is far from the first time that labor leaders have been at pains to state that they aren’t against technology for fear of being painted as opposing progress itself, but it may well be a pipe dream to imagine that workers can influence AI to have a liberatory rather than deleterious effect on our lives. Even bracketing the technology’s risks beyond the workplace, wherever one looks, AI is wielded as a means of controlling workers, managing and surveilling them, and yes, maybe one day, replacing them. This technology is not “introduced” to the workplace; it is imposed.
As unionized workers fight to secure defensive measures against such managerial impositions in their contracts, they must use those experiences as a springboard and testing grounds for legislative and regulatory fixes. If AI threatens a mind-boggling array of workers, it also unites them, offering a shared enemy against which the labor movement can and should face down. That can’t happen a moment too soon: nonunion workers are depending on it.