Could AI be used to replace striking workers?

When New York Times tech workers went on strike the day before the election, the newspaper’s publisher, AG Sulzberger, criticized the move, saying it was “troubling that the Tech Guild would try to block [the paper’s] public service at such a consequential moment for our country.”

Some of the striking workers handle software and data analysis. It wasn’t clear if, without them, the paper’s website would be able to handle what was an expected influx of election-related traffic. Aravind Srinivas, CEO of AI company Perplexity, responded on X to Sulzberger’s statements, saying that his company was “on standby to help ensure your essential coverage is available to all through the election.”

The tech workers’ strike ended after a week (though without a contract resolution), and there were no reported website outages. But the offer from Srinivas struck many as a way to undercut the union’s power and compromise workers’ ability to fight for better labor conditions. Replies to his comment called him a “scab” (a term for someone who crosses a picket line and replaces striking workers).

Perplexity did not respond to a request for comment from Fast Company, but Srinivas did say on X that the offer was “*not* to ‘replace’ journalists or engineers with AI but to provide technical [infrastructure] on a high-traffic day.” (In fact, that’s the tech workers’ jobs—to provide that infrastructure—not journalists or engineers who, of course, were not on strike.)

Though it’s not clear what work Srinivas was offering his AI services to do, the situation highlighted the potential ways in which AI could undercut labor actions in general—and the concerns workers have about how their employers may be integrating such tools.

Unions and AI

Sharon Block, executive director of the Center for Labor and a Just Economy at Harvard Law School, says the offer from Perplexity—even though the Times didn’t take the company up on it—“just underscores how important it is for people to have a union to be able to help them navigate this time.”

Though union participation is still low in the U.S. (only about 6% of private sector workers are in one), public support for them has been increasing. A 2022 Gallup poll found that 71% of Americans support unionization efforts, a record high since 1965. And as the labor union has grown in the past few years, more and more workers have turned to strikes to exercise their power. In 2023, some 539,000 U.S. workers went on strike—a 141% increase compared to 2022.

That number was bolstered by actions including the Hollywood writers’ and actors’ strikes, both of which touched on the use of AI. Workers won guardrails around AI use and writer credits, including that a company cannot require a writer to use AI, and protections around actors’ digital likenesses. “We saw in the Hollywood strikes, it’s not about giving workers a veto over progress,” Block says. “It’s about giving people a voice in really important issues about how work gets done.” And if a worker isn’t in a union, she adds, then they don’t get the opportunity to be a part of that conversation.

Labor unions are just beginning to negotiate over the use of AI, says Virginia Doellgast, an employment relations and dispute resolution professor at Cornell’s School of Industrial and Labor Relations. And the most active areas have been around creative workers through the Writers Guild or SAG-AFTRA. That sort of work is unique because there are issues around ownership rights and the use of AI. But it’s harder to protect the work of people not in creative industries, and AI is already replacing some more technical jobs such as data entry, she notes.

AI and picket lines

In the case of the New York Times tech workers strike, the use of AI to keep the website running is a bit complicated. Workers were striking to protest unfair labor practices committed by the employer. That’s different than an economic strike, the point of which is to push for a concession like higher wages. During an economic strike, it’s legal for employers to permanently replace workers. But with an unfair labor practice strike, they can’t.

Employers also aren’t allowed to make “unilateral changes” to the terms of conditions of someone’s employment, once their union is certified. But how AI fits into all this isn’t yet quite clear. “Would using AI be a change to their employment?” Block asks of the tech workers’ situation. “Would it be like hiring strike replacements?” It’s not yet clear how labor law will treat AI.

To Block, the offer from Perplexity sounded more akin to hiring replacements. (Temporary replacements are allowed in unfair labor practice strikes, but workers must be allowed to return to their jobs at the end of the action.) Still, there might not be a consensus on that view. “We’re in a brave new world,” she says. “That’s why it’s better when people actually have the opportunity to sit down and think through these things, and bargain over them.” It’s an issue labor law will likely have to contend with in the future.

Though the use of AI here might be new, the overall issue—of employers’ ability to fight labor actions—isn’t. “If you think historically, you don’t need AI to cross picket lines [in order] to weaken a strike or weaken resistance,” Doellgast says. “There are many different ways that employers threaten to outsource jobs or threaten to downsize work.”

Presumably, employers could threaten to outsource a job to AI, but that threat is only credible, she adds, “if that AI is significantly cheaper and as effective as the workers themselves,” and she doesn’t think the technology has proven itself yet.

When speaking to software engineers for her research, some told her that AI tools help their programming work, but others said it’s like “consulting tarot cards—it’s not accurate enough to actually do a good job at helping me to write code and to solve problems.” Even Big Tech companies seem to acknowledge this; when Google launched new AI tools for Chrome, a pop-up warned, “This feature uses AI and won’t always get it right.” Employees who work with AI tools often mention having to check its work.

That means workers aren’t yet facing the threat of being completely replaced by AI “in one fell swoop,” Doellgast says. But there are still issues to work out about how AI is being integrated into everyday work. She’s heard of situations where managers have put AI tools in place without much thought, and have run into problems. That again shows the need for workers to be involved in these discussions, and that workers can make AI models more effective and AI tools more efficient.

“If you’re in a situation of conflict with your workers, you really need those workers to work with the AI—to train models, to correct mistakes, to use it in a way that’s sort of supplementing or complementing worker skills—rather than just purely replacing them,” she says. Using AI as a threat to replace a job “is not necessarily a winning strategy in the longer run for companies or for employers,” she adds. Doing so would likely only antagonize a union even more.