Workers Are On Their Own When It Comes to AI Regulation
Contract negotiations (and unions overall) are one of the best tools workers can use to challenge standards of work: with federal regulation pointedly looking the other way, it might be our best bet.
Last week Biden's administration released a "bill of rights" about AI regulation, presumably connected to OpenAI CEO Sam Altman's appearance in front of Congress.
This “Bill of Rights” mostly focuses on challenging algorithmic bias and federal use of AI. The federal agencies named (including the FTC) are equipped to regulate questions related to those topics in particular — not that the Bill of Rights regulates anything, to be clear. It’s just a list of stuff that could be good to think about at some point, haha, if you guys in Big Tech are interested, haha, you know? I don’t know, we were just thinking about it — but you guys can decide!
Relative to the power of “being the President” — this Bill of Rights is equivalent to the PDFs I used to make in Canva when my nonprofit ED bosses would email me at 9 PM “are we saying something about this?!” There’s not a lot there in terms of power. As Wired said, the bill was a toothless white paper.
To be fair, data policy is also an artificial intelligence policy. What is in this bill of rights addresses AI’s critical implications for data privacy, IP, and regulation.
Still — that there’s no reference to workers, or automation, even in a disingenuous Target Pride Month kind of way, in that cross-agency bill of rights at all.
That radio silence isn’t because AI hasn’t yet changed work: already CEOs are openly admitting that AI has already transformed their retention and hiring.
Weirdly, tech companies have been more vocal about the need for AI regulation than the government has, with everyone from Musk to Altman himself espousing the need for regulation, desperately hoping the government will create some kind of agency to regulate AI (which is very different from creating an effective agency to regulate AI). Big tech’s desire for regulation is kind of bogus, but at least they’ve addressed it — but in the federal government, it’s quiet when it comes to AI and labor.
It’s not surprising that politicians aren’t bringing up labor much in the AI policy conversations. Biden hasn't been much of a labor president (in spite of his claims).
Some US politicians (like Chuck Schumer) have contended there's little to be done when it comes to AI until it' “fully develops.” But without much action (or even fake grandstanding about taking action) those quotes sound wildly out of their depth, similar to the questions like “Does TikTok connect to WiFi?” in the TikTok Congressional Hearing.
This is especially true because spite of what national policymakers are saying, there's plenty that can be done when it comes to regulating AI in the workforce.
We know this because of the WGA Writer’s Strike.
AI and Labor in The WGA Writer’s Strike
Many people have written far more eloquently than I will about the context and stakes of the WGA Writer’s Strike, from issues with residuals to the size of writers’ rooms. AI is a major part of the WGA’s negotiations, in part because the technology presents an existential threat to the union. It’s likely that the final outcome of negotiations will hinge on regulations around its use by studios.
As writer and member of the guild negotiating committee James Schamus wrote in The Guardian, the problem with AI has always been about how it will undoubtedly be used by studios — to cut costs, hoard profit, and “transform jobs into the endless, frantic scramble for gigs.”
Here are the WGA’s main demands about AI:
Writers who are working on adaptations or starting with “source material” are paid less by current standards. The WGA argues that AI-generated material would not be considered “source material” under the union’s contract. This rule would make sure that studios can't pay writers less for using AI-generated material.
AI shouldn't be allowed to write on its own. Studios wouldn't be allowed to let AI programs create scripts by themselves or edit scripts written by human writers.
Most AI is powered by huge datasets without much discretion re: copyright, IP, or anything else related to who owns that data. (These are part of the issues that Biden’s AI Bill of Rights does address). The WGA contends that studios cannot use AI that includes the work of members in the union without paying some kind of fee.
As these potential regulations highlight, the issues with AI and labor are about who gets paid, not about AI itself.
In response to the WGA’s demands, the AMPTP (the body that collectively represents studios in negotiations) hasn’t said much about in terms of their intentions with AI. In an early conversation, AMPTP suggested “annual meetings to review new technology” as a counter to the WGA’s AI proposal. An annual meeting isn’t quite a pizza party, but it’s close.
And the contrast between the awkward noncommittal takes from national policymakers and how the Guild is approaching AI in their contract negotiations makes it really clear: negotiations like these where workers have real power, are really the only way to change anything about AI and work for the benefit of workers.
Beyond the WGA Strike
Contract negotiations across industries have always been about, to some extent, who is allowed to do what when it comes to work.
For example, when the IBEW creates regulations around green manufacturing, that in turn impacts wages, safety, and support across the whole industry. It’s how trade unions, for example, who historically have not been great when it comes to uh, inclusion, have still been able to set wage standards + workforce requirements in a way that changes the entire labor market.
Those in labor are probably thinking — duh! By organizing collectively, workers can attempt to secure better conditions and rights, leading to a more equitable distribution of wealth and power within society. Duh!
But I’m not in a trade union, are you? As Hamilton Nolan pointed out recently, less than 6% of the private workforce is unionized. This number is even smaller in tech companies (where AI replacement is happening fastest).
It’s unlikely (barring a sudden labor explosion in the industry) that mass workplace contract negotiations are going to meaningfully change what AI is allowed to do in white-collar industries.
But as this extremely cool Teamster with a Jimmy Hoffa tattoo said in an interview, the WGA’s fight to stop AI is bound up with all workers’ ability to define what AI can be used for and when.
The actual, concrete demands around AI and its usage in Writer’s Rooms from the WGA demonstrate how workers, can actually get to decide the terms of their work, including when and how new technology is used.
And as workplace strikes and unions grow exponentially, particularly in the service industry (where food and retail workplaces like Sweetgreen that once touted their great worker benefits are now considering 'automating salad'), there is a major opportunity for workers to learn from the WGA, no matter what the outcome of the Strike, and make AI more trouble than it’s worth for employers in every contract negotiation.
Contract negotiations (and unions overall) are one of the best tools workers can use to challenge the power dynamics between them and their bosses. With the federal government acting the way it does, when it comes to AI, those negotiations might be the only tool left.