Intro
Here's a closer look at where that fight is playing out, profession by profession: the specific deals, lawsuits, and statutes each one has produced so far.
Table of Contents (jump links)
Screenwriters and Actors · Radiologists · Lawyers and Paralegals · Court Reporters · Truck Drivers · Translators and Interpreters · Teachers · Musicians · Financial Analysts and Accountants · Customer Service Workers
Journalists and Newsroom Workers · Authors and Book Publishing · Illustrators, Photographers, and Visual Artists · Dockworkers and Port Workers · Airline Pilots · Warehouse and Logistics Workers · Gig and App-Based Drivers · Fashion Models · Insurance Claims Reviewers · Content Moderators and AI Raters
Screenwriters and Actors
Hollywood's 2023 labor dispute was really two overlapping strikes — the Writers Guild of America from May 2 to September 27, and SAG-AFTRA from July 14 to November 9 — the first time in more than sixty years that writers and actors had walked out together. AI sat alongside streaming pay as the central issue in both fights, and each union came away with a different, highly specific answer to it.
The WGA's contract, ratified in October 2023, drew a bright line: generative AI cannot be credited as a writer, and AI-generated material can't count as "literary material" or "source material" that a human writer is merely revising — a distinction that matters because writers are paid less to rewrite someone else's draft than to originate their own. Writers can choose to use AI tools if they want to, but a studio can't require it, and any company that hands a writer AI-generated pages has to say so upfront. The guild also reserved the right to argue that training AI on its members' produced scripts without permission violates the agreement.
SAG-AFTRA's contract, ratified two months later with 78 percent member approval, went further into technical detail, creating separate rules for what it calls "Employment-Based" and "Independently Created" digital replicas. A studio that wants to scan a performer's face or clone their voice now needs consent that is clear, conspicuous, and separately signed, not buried in the standard paperwork, and has to describe exactly how and where the replica will be used, episode by episode, with new consent required whenever that changes. Digital replicas also can't be used to avoid hiring background actors for scenes that would otherwise need them.
The fight didn't end at ratification. Video game voice and motion-capture performers, covered by a separate SAG-AFTRA contract, walked out in 2024 over the identical issue and didn't settle until July 2025, ending a nearly year-long strike with an Interactive Media Agreement that added the same consent and disclosure protections and let performers suspend consent for new AI material during any future strike. The union kept building on multiple fronts after that: it helped draft the federal NO FAKES Act, introduced in the Senate in mid-2024 with the unusual backing of the Motion Picture Association, the RIAA, and OpenAI alike; it struck separate voice-licensing deals with AI companies including Replica Studios and Narrativ that guarantee per-use payment; and in mid-2025 it helped beat back a provision in a federal tax bill that would have barred states from regulating AI at all.
Radiologists
In 2016, AI pioneer Geoffrey Hinton told an audience that hospitals should stop training radiologists altogether, arguing it was "completely obvious" that deep learning would out-diagnose them within five years, ten at the outside, and that medicine already had all the radiologists it would ever need. It became one of the most quoted predictions in the entire history of AI-and-jobs discourse, and it set off years of anxiety among radiology residents who wondered if they were training for a job that wouldn't exist by the time they finished.
Nearly a decade later, the prediction has landed almost exactly backwards. The Mayo Clinic's radiology staff grew 55 percent between 2016 and today; the American College of Radiology now forecasts the specialty's physician supply will grow another 26 percent over the next thirty years; and radiology describes itself as facing a historic shortage, with imaging backlogged for months at some centers. Hinton has since said publicly that he got the timeline wrong, though he still expects the interpretation of scans to eventually be split between radiologists and AI rather than handled by either alone.
What actually happened is that the specialty organized around a narrower, more defensible position instead of resisting AI outright: a licensed physician stays the final decision-maker of record, full stop. The American College of Radiology built real infrastructure to make that stick: Assess-AI, a national quality registry launched in late 2024 that tracks how imaging algorithms perform in day-to-day practice rather than in a vendor's marketing materials, and, in May 2026, the first-ever ACR-SIIM Practice Parameter for Imaging AI, a formal clinical standard governing how these tools get validated and monitored once deployed. The college has also lobbied regulators for physician-specific payment codes tied to AI-assisted reads, on the theory that if reimbursement flows to the algorithm alone, the financial incentive to keep a radiologist in the loop disappears along with it.
Lawyers and Paralegals
The profession's cautionary tale dates to Mata v. Avianca in 2023, when two New York lawyers were sanctioned for filing a brief built on ChatGPT-generated citations to cases that turned out not to exist. Most observers assumed that would be a one-time embarrassment the profession would learn from quickly. Instead it became a genre: a tracker maintained by legal researchers had logged more than 600 cases nationwide by late 2025 in which a lawyer cited fabricated AI-generated authority, more than 50 of them in California alone.
The sanctions have kept coming, and the dollar amounts have grown. In February 2025, a federal court fined three attorneys at the national firm Morgan & Morgan for citing fabricated cases pulled from the firm's own in-house AI research tool. In October 2025, a California appellate court fined a solo practitioner $10,000 over a brief containing 21 fake citations generated by ChatGPT, and in the process considered, for what may be the first time, whether opposing counsel has a duty to catch an adversary's AI-fabricated citations. Even emergency immigration filings haven't been spared: a federal court sanctioned an attorney in July 2025 for submitting fabricated quotations generated by Anthropic's Claude in an emergency deportation case, rejecting the argument that a tight deadline and the attorney's own illness excused the failure to verify.
Bar associations have responded with rules rather than outright bans. Guidance issued by state bars makes clear that AI cannot give legal advice directly to a client, and that a licensed attorney has to personally review and take responsibility for anything the software produces; one sanctioning court put it plainly, ruling that the duty to check sources doesn't change just because the drafting tool did. California's Judicial Council moved in late 2025 to formally regulate AI use in the courts. None of this amounts to banning AI from legal work; a 2024 Stanford RegLab study found three out of four lawyers already planned to use generative AI in their practice. It amounts to insisting that whatever the software gets wrong, a human with a license and something to lose is still the one who signed the filing.
Court Reporters
Court reporting is the one profession here where the AI threat and a pre-existing labor shortage are tangled together, which makes the politics messier than a simple story of robots replacing workers. Stenography takes years to learn; certified reporters write in machine shorthand at 225 to 300 words a minute, and fewer people are entering the training pipeline each year while experienced reporters retire. California alone was short 458 full-time certified court reporters as of January 2025, and nearly half of the state's active licenses had been issued more than thirty years earlier.
That shortage, not simple cost-cutting, is what's actually opening the door to digital recording and AI transcription. But certified reporters have fought to keep the door from opening all the way. Federal Rule of Civil Procedure 80 requires that a stenographically reported transcript be certified by the person who took it down before it can be used as evidence at a later trial; a certified human being, not a device, is what makes a transcript legally provable on appeal. The National Court Reporters Association leans hard on that distinction, and on cautionary tales: in the high-profile 2023 murder trial of Alex Murdaugh, a backup digital recording of the proceedings turned out so flawed it was useless to the defense.
The legislative compromise taking shape looks less like resistance than managed coexistence. California's Assembly Bill 1189, moving through the legislature in 2026, would create an official certification track for digital court reporters and transcriptionists to help close the staffing gap, while explicitly barring courts from firing stenographers just to replace them with cheaper digital alternatives. Kentucky has used digital recording as its official court record since the 1980s, proof the two systems can coexist. But more than half of U.S. states still restrict who's legally allowed to produce an official transcript, protecting the certification itself even while scrambling to fill the shortage that certification helped create.
Truck Drivers
No profession on this list has fought AI more directly through state legislatures than trucking. Since 2023, the Teamsters have backed "driver-in" bills — measures requiring a licensed human operator behind the wheel of any autonomous commercial truck — in at least a dozen states, including California, Colorado, Maryland, New York, Iowa, Indiana, and Alaska. California Governor Gavin Newsom vetoed versions of the bill in both 2023 and 2024, arguing existing regulators were better positioned to handle safety; a version passed the state Assembly again in 2025, even as California's DMV moved the opposite direction in May 2026, adopting rules that let autonomous vehicle companies apply for permits to test and deploy heavy-duty self-driving trucks on public roads.
Colorado's fight escalated furthest: after the bill stalled in the legislature, Teamsters gathered support for House Bill 26-1286, which would let voters decide directly, in the state's first-ever ballot referendum on autonomous vehicle rules, whether trucks over 26,000 pounds need a human in the cab. The union has paired state campaigns with a federal one, publishing a five-point framework asking Congress to regulate the vehicle, the operator, and the operations of autonomous trucks the way it already regulates human-driven ones, and to let states keep rules stricter than whatever floor Washington eventually sets.
The union points to real-world stumbles as evidence its caution is warranted: Waymo and Zoox both issued recalls after safety incidents in 2025 and 2026, and Aurora Innovation, one of the most closely watched driverless-trucking startups, quietly put human safety operators back into its trucks after less than three weeks of driving without them. Critics counter that the broader safety data cuts the other way: Waymo's driverless fleet had logged 170 million miles without a human behind the wheel through December 2025, and groups like the Goldwater Institute argue driver-in mandates are safety arguments in name only, designed to protect an existing labor model rather than the traveling public. Either way, trucking is the clearest example here of a profession trying to write its own job security directly into statute.
Translators and Interpreters
Immigration courts draw the clearest legal line in this fight. Under federal regulation, every document filed with an immigration court must either be in English or come with a certified translation, and the certifying translator has to personally affirm, on the record, that they are competent in the relevant language and that the translation is accurate — an affirmation a machine cannot legally make. Federal courts, state courts, and agencies like USCIS impose similar requirements for anything that will actually be relied on in a proceeding.
In practice, the industry has settled into a two-tier system rather than a ban on AI. Translation companies now routinely use machine translation to triage high-volume, low-risk material, sorting discovery documents, giving a lawyer a rough sense of what a foreign-language file says, while reserving certified human translators for anything that will actually be filed with a court or an immigration judge. Industry research has found that legal translations, even human ones, carry meaningful error rates in grammar, vocabulary, and formatting, which certified translators point to as evidence of how much can go wrong once there's no accountable person attached to the final product.
Court interpreting has followed a parallel path. Becoming a certified court interpreter in states like California requires passing both a written exam and an oral bilingual interpreting exam, and courts have generally treated AI transcription and translation as an aid to understanding rather than a substitute for a sworn interpreter in the room. The professional argument is less about whether AI is fluent enough — even advocates for certification concede it often is — than about who is legally and professionally on the hook when a mistranslated line of testimony or a garbled asylum declaration changes the outcome of someone's case. In the settings that matter most, the profession has made sure the answer still has to be a certified human being.
Teachers
Teachers' unions have arguably shifted strategy more visibly than any other profession on this list. In a 2023 resolution, the American Federation of Teachers declared that AI's effect on the workplace is a mandatory subject of bargaining and committed to writing contract language limiting how it could be used to displace educators. In 2024, it leaned toward engagement rather than confrontation, putting more than $200,000 into an Innovation Fund that helped eleven school districts train teachers on AI and stand up teacher-led committees to shape local ed-tech policy.
That engagement culminated in July 2025 with the National Academy for AI Instruction, a $23 million partnership between the AFT and Microsoft, OpenAI, and Anthropic to train the union's 1.8 million members on AI — a deal AFT president Randi Weingarten defended as putting teachers in the driver's seat rather than leaving AI policy entirely to school boards and tech vendors, while critics, including some of the union's own members, accused her of legitimizing tools that were already reshaping the profession without teachers' input. The National Education Association took a more conventional labor-relations approach at its 2025 convention, voting to develop model contract language covering job protections, data privacy, and AI-literacy training that local affiliates could adapt at their own bargaining tables.
By May 2026, the AFT's posture had hardened again. Weingarten called for large-scale rollbacks of student AI and device use in classrooms, new restrictions on AI tools aimed at younger students, and a "Big Tech Tax" to fund the research and training she argued companies had offloaded onto already-stretched school districts, describing the prior two years of unregulated ed-tech adoption as an uncontrolled experiment run on children. Reporting on the union's own contracts has found that despite all this activity, only a handful of the country's school districts have actually written AI-specific protections into a ratified labor agreement, leaving most teachers, for now, governed by whatever policy their district sets unilaterally rather than anything actually bargained.
Musicians
The music industry's fight has played out mostly in federal court rather than at a bargaining table. In June 2024, the RIAA sued the AI platforms Suno and Udio on behalf of the major labels, accusing both of training their models on copyrighted recordings without permission or payment and of generating output that could closely mimic specific vocal styles and hit songs; the labels sought damages of up to $150,000 per infringed work.
By late 2025, the litigation had split along an interesting fault line. Warner Music settled with Suno in November 2025 and converted the relationship into a licensing partnership, with Suno agreeing to phase out the models it had trained without permission in favor of licensed versions. Universal Music settled with Udio a few weeks earlier, in October 2025, securing a per-generation royalty of roughly a fifth of a cent to half a cent per track and a jointly licensed AI music platform planned for 2026. Sony Music refused to settle and, as of mid-2026, is still litigating against both companies, with a closely watched fair-use ruling in the Suno case expected from a Massachusetts federal court.
Working musicians without a major label's legal budget have filed their own, separate class actions against Suno and Udio, arguing independent artists deserve the same licensing leverage the majors negotiated for themselves. The argument underneath all of it is the one session musicians, composers, and touring artists have been making since the lawsuits began: a voice or a style built from decades of recordings is still somebody's labor, and if it isn't treated as something owned, there's nothing left to license or negotiate over. Even Suno, while still fighting in court, now says publicly that transparency and artist compensation matter — a sign of how far the ground has shifted since 2024, when training on unlicensed recordings without comment was still a viable business strategy.
Financial Analysts and Accountants
Accounting has embraced AI more openly than almost any profession on this list, which makes its version of resistance look less like a fight and more like a licensing strategy. The 2025 AICPA and CPA.com AI in Accounting Report describes firms rolling out "agentic" AI that can categorize transactions, flag anomalies, draft client reports, and generate real-time forecasts largely on its own — exactly the kind of first-pass bookkeeping and analysis work that used to occupy junior staff for entire days.
What hasn't moved is who is allowed to sign their name to the result. A CPA license is still required to sign an audit opinion, and that license is still what regulators can revoke if the opinion turns out to be wrong, meaning no matter how much of the underlying work an algorithm does, a named, licensed human remains the accountable party on anything that carries legal weight. The AICPA's new Quality Management Standards, which firms had to implement by December 2025, reinforce that a human reviewer has to stay in the loop even as the workflow around them automates.
The profession has gone further than most by trying to turn oversight of AI into a new line of work rather than just a defensive position. In 2025, ISACA launched an Advanced in AI Audit certification specifically for auditors who already hold a CPA, CIA, or CISA credential, built on the theory that accountants are uniquely positioned, through training in evidence, controls, and professional skepticism, to become the people who audit AI systems themselves, not merely the people who use them. The profession's own trade press captured the sentiment in an October 2024 piece arguing that AI would make the job more efficient without eliminating the need for an auditor to do it.
Customer Service Workers
Roughly three million Americans work in call centers, and the sector has become one of organized labor's clearest AI battlegrounds precisely because so much of it was already low-margin, closely monitored, and easy to relocate. The Communications Workers of America has made AI protections a bargaining priority across every industry where it represents workers — call centers, telecom, airlines, journalism — leaning on the leverage of an enforceable contract rather than waiting on legislation that may never arrive.
Some of those contracts have produced real, specific protections. Las Vegas casino workers represented by the Culinary Workers Union won a 2023 contract guaranteeing severance of $2,000 per year worked for any employee whose position is eliminated by technology or AI, along with a requirement that management bargain before rolling out any new AI-driven system. In May 2026, CWA members at the National Abortion Federation's hotline staged a 24-hour unfair labor practice strike after management refused even minimal AI protections, and won a three-year contract explicitly barring unilateral AI implementation. The International Longshoremen's Association negotiated what may be the strongest protection on this entire list: a complete prohibition on fully automated port technology.
Legislators have joined in on a narrower but related front. A bipartisan bill introduced in 2025 by Senators Ruben Gallego and Jim Justice, the Keep Call Centers in America Act, would require companies to notify the Department of Labor 120 days before moving call center jobs overseas, in what the CWA described as a response to companies using AI to speed up and de-skill work domestically while offshoring it abroad at the same time. None of this stops a company from deploying a chatbot. What it increasingly does is guarantee that a customer who wants one, and a worker whose job depends on it, both get some say in when a human stays on the line.
Journalists and Newsroom Workers
The news industry's fight over AI has unfolded on two fronts at once: the bargaining table and the courtroom. The New York Times sued OpenAI and Microsoft for copyright infringement in December 2023, seeking billions of dollars over the use of Times journalism to train ChatGPT; more than two years later the case remains in discovery in the Southern District of New York, with fights over access to ChatGPT conversation logs and no trial date set. The News/Media Alliance, representing more than 2,000 outlets, has backed the suit, while other publishers have instead signed licensing deals with AI companies rather than litigate.
Where journalists have more direct leverage is inside their own newsrooms, and NewsGuild-CWA locals have used it aggressively. More than fifty newsroom contracts now include AI-specific language. At POLITICO and E&E News, the first contract, ratified in 2024, required 60 days' notice and good-faith bargaining before management could roll out AI tools that materially changed job duties or risked layoffs, and in December 2025, the PEN Guild won a landmark arbitration ruling that POLITICO had violated exactly that provision. Workers at the New York Times Tech Guild struck for eight days to win a semi-annual joint labor-management committee on generative AI, and other locals, from the New Republic to newsrooms across the Gannett chain, have bargained similar disclosure and consultation requirements into new contracts.
The union has also taken the fight public and legislative. In December 2025, NewsGuild-CWA launched a national campaign against what members call "AI slop" — AI-generated or AI-assisted content published without adequate editorial oversight — and it is backing New York's proposed FAIR News Act, which would require news organizations to disclose AI use to both workers and readers. None of this stops outlets from adopting AI tools, and many already have; what it increasingly guarantees is that journalists get advance notice and a seat at the table before those tools reshape their jobs.
Authors and Book Publishing
Book publishing's confrontation with AI has played out almost entirely in federal court, and by mid-2026 it had already produced the largest copyright class-action settlement in American history. Authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson sued Anthropic in 2023 over its use of pirated books, downloaded in bulk from sites like Library Genesis, to train its Claude models. A federal judge ruled in mid-2025 that training on legally acquired books was protected fair use, but that using pirated copies was not, clearing the way for a class covering roughly 480,000 works to seek damages over the piracy itself. Facing potential statutory damages that could have run into the tens of billions of dollars, Anthropic agreed to a $1.5 billion settlement in September 2025; the court granted final approval on July 20, 2026, with authors and publishers receiving a little over $3,000 per work on average and attorneys awarded roughly $100 million, well below the $300 million originally requested.
A parallel and still-unresolved fight is underway against OpenAI. The Authors Guild, joined by authors including George R.R. Martin, John Grisham, and Jodi Picoult, sued OpenAI in 2023 over training ChatGPT on their books. In October 2025, a federal judge allowed a new theory to proceed, ruling that ChatGPT's own AI-generated plot summaries — in one instance closely mirroring Martin's writing — could themselves constitute copyright infringement, distinct from the underlying training question. The case remains active, with a bellwether trial expected in late 2026.
Short of litigation, the Authors Guild has pushed publishers toward contractual protection. Its Model Contract Clauses, released in 2026, have become a reference point for authors and agents negotiating AI-rights language into new book deals, and its "Created by Humans" program offers writers a way to certify and market their work as free of AI generation.
Illustrators, Photographers, and Visual Artists
Visual artists were among the first creative professionals to sue over AI image generation, and more than three years later, their signature case is still unresolved. Cartoonist Sarah Andersen, along with Kelly McKernan and Karla Ortiz, sued Stability AI, Midjourney, and DeviantArt in January 2023, arguing that training Stable Diffusion on billions of images scraped from the internet, including their own, infringed their copyrights and let users generate new work "in the style of" specific living artists. A federal judge dismissed most of the original claims later that year, but let a core infringement claim against Stability AI move into discovery; by early 2026 the artists had filed a third amended complaint, adding the video-AI company Runway as a defendant, with the case still years from trial.
Getty Images pursued a parallel, far larger case in the UK, and came away with little. Getty had dropped its central claim, that training Stable Diffusion on its images was itself UK copyright infringement, before trial, for lack of evidence the training happened in Britain. In a November 2025 ruling described as the first major British test of AI and copyright law, the High Court then rejected Getty's fallback argument that merely importing the already-trained model into the UK was infringing, and it dismissed Getty's database-rights claim entirely; Getty won only a narrow trademark claim, based on its watermark occasionally surfacing in Stable Diffusion's output. A separate, related Getty suit in U.S. court continues.
Meanwhile, the U.S. Copyright Office has held its ground on a narrower but foundational question: whether AI output can be copyrighted at all. Its January 2025 report reaffirmed that copyright protects only human authorship, concluding that typing prompts into an AI system, however detailed, does not make the user an "author" of what comes out. That position survived a direct legal challenge in March 2025, when the D.C. Circuit Court of Appeals ruled against inventor Stephen Thaler, who had argued that a work generated autonomously by his AI system deserved copyright protection in its own right. Photographers, illustrators, and stock-image agencies have since leaned on that principle to argue that AI-generated competition undercuts a market built on work only a human can actually own.
Dockworkers and Port Workers
No profession on this list backed its resistance to automation with more disruptive leverage than the longshore industry. The International Longshoremen's Association, representing 85,000 dockworkers at ports along the U.S. East and Gulf Coasts, walked out for three days in October 2024 in a strike estimated to have cost the U.S. economy roughly $2 billion, specifically over the United States Maritime Alliance's refusal to rule out expanding automated and semi-automated equipment at container terminals.
A tentative deal reached on January 9, 2025, days ahead of a new strike deadline, headed off a second walkout. ILA members ratified the resulting six-year master contract on February 25, 2025, with nearly 99 percent in favor; the agreement, formally signed that March, runs through September 2030 and delivers a 62 percent wage increase alongside what ILA President Harold Daggett called "iron-clad" automation protections — language restricting new automated equipment while allowing limited technology in exchange for job guarantees. Daggett called it the "gold standard" for dockworker contracts anywhere in the world.
The union didn't stop at its own contract. In November 2025, more than a thousand dockworkers and maritime union leaders from around the world gathered in Lisbon for an "Anti-Automation Conference," signing a document establishing a Global Maritime Alliance aimed at coordinating opposition to port automation on every continent, an explicit attempt to keep employers from playing automation-friendly ports in one country against protected ones elsewhere. Labor attorneys who tracked the U.S. negotiations don't expect the underlying tension to disappear, predicting technology and job security will remain a flashpoint in every future round of bargaining, however strong any single contract's protections look on paper.
Airline Pilots
Commercial aviation's version of this fight is preemptive: pilots are organizing against a change regulators haven't approved yet, on the theory that waiting for a formal proposal would already be too late to stop it. The Air Line Pilots Association, the world's largest pilot union, launched its "Safety Starts with 2" campaign in June 2023 to oppose any reduction below two pilots on the flight deck, whether marketed as reduced-crew operations, "extended minimum crew operations" (eMCO), or full single-pilot operations. Its most immediate target has been Europe, where aircraft maker Airbus and the European Union Aviation Safety Agency have been developing eMCO standards, with early talk of an Airbus A350 flying with a reduced crew as soon as 2026.
That timeline has since slipped, which ALPA credits to sustained pressure from pilot unions on both sides of the Atlantic. EASA's leadership has acknowledged it cannot yet demonstrate an equivalent level of safety for reduced-crew flights compared to today's two-pilot standard, and its most recent rulemaking update shifted toward developing flight-deck technology first rather than certifying reduced crews outright. ALPA describes the shift as a pause rather than a victory, expecting manufacturers and airlines to revive the push as cost pressures continue.
The union has also taken the fight to Washington. In July 2025, Senators Tim Sheehy and Tammy Duckworth led a bipartisan letter, joined by 43 colleagues, urging the Department of Transportation and FAA to formally oppose single-pilot proposals at the International Civil Aviation Organization, warning that eMCO would leave a lone pilot in sole command of a cruising airliner while a second rests elsewhere on the plane. ALPA has paired that outreach with a simple public message: automation can assist a pilot, but it should never get to replace one.
Warehouse and Logistics Workers
Warehouse automation predates the current AI boom, but the fight over it has accelerated alongside it. California passed the first significant law in the space, AB 701, in 2021, requiring large warehouse employers to disclose written productivity quotas to workers and barring quotas that make it impossible to take a legally required bathroom, meal, or rest break, a direct response to reports that algorithmically set pace requirements at large e-commerce operators were driving injury rates well above industry averages.
The fight has since moved from disclosure toward control over the algorithm itself. California's "No Robo Bosses Act" (SB 7), introduced by state Senator Jerry McNerney and backed by the California Federation of Labor Unions, would have required meaningful human oversight before an automated system could discipline or fire a worker; it passed both chambers of the legislature in 2025 but was vetoed by Governor Gavin Newsom that October, who argued it overlapped with new automated-decision-making rules already being finalized by the state's privacy regulator. McNerney reintroduced a revised version, SB 947, in February 2026, addressing the governor's stated objections.
Organized labor has pursued the same fight directly with employers, most visibly at Amazon. Workers at the company's JFK8 warehouse on Staten Island unionized in 2022 and affiliated with the Teamsters as ALU-IBT Local 1 in 2024; after what the union describes as years of deliberate delay, the National Labor Relations Board ordered Amazon to begin bargaining with the local in spring 2026. Teamsters General President Sean O'Brien has long called Amazon an "existential threat" to the union's core warehousing and trucking membership, a concern sharpened by leaked internal planning documents reported by the New York Times describing a company goal of automating up to 75 percent of fulfillment operations. Amazon has pushed back publicly on the idea that automation means fewer jobs, pointing to continued hiring even as it rolls out robotic systems like its newly unveiled "Blue Jay" picking robot across a growing number of warehouses.
Gig and App-Based Drivers
Rideshare and delivery drivers face a different kind of automation threat than most professions on this list: the algorithm isn't replacing them outright, it's managing them, setting their pay, assigning their trips, and deciding, often without explanation, when to cut off their income entirely. A Washington state law gave app-based drivers a right most gig workers nationwide still lack: if deactivated for more than three days, a driver can formally appeal and seek compensation for lost wages, and platforms must show "just cause" for the decision rather than simply citing an algorithmic score.
Colorado followed in June 2024, when Governor Jared Polis signed HB24-1129 and SB24-75 after organizing by Colorado Independent Drivers United, requiring companies like Uber, Lyft, and DoorDash to give drivers greater transparency into how fares, routes, and deactivation decisions are calculated. Momentum has since moved to Congress: the Empowering App-Based Workers Act, introduced in the Senate in July 2025 by Senators Brian Schatz and Chris Murphy and in the House that December by Representatives Pramila Jayapal, Donald Norcross, and Ilhan Omar, would require detailed pay statements, guarantee drivers at least 75 percent of each fare, and force platforms to disclose how their algorithms make pay, assignment, and discipline decisions. Los Deliveristas Unidos, the New York City-based delivery-worker campaign that already helped win one of the nation's first minimum-pay standards for app-based delivery workers, has championed the bill nationally.
The stakes are more than theoretical. A 2025 Human Rights Watch survey found that more than a third of gig workers had been deactivated from a platform at least once, and that close to half of those deactivations were later determined to be mistakes, errors that, absent a law like Washington's, can end a driver's income with no appeal and no explanation beyond whatever the app's support chat is willing to offer.
Fashion Models
Fashion may be the newest profession to win AI-specific legal protection, and New York moved on it faster than almost anywhere else. The New York Fashion Workers Act, effective June 19, 2025, requires model management companies and their clients — retailers, designers, ad agencies, photographers — to get a model's clear, separate, written consent before creating or using a "digital replica": an AI-generated or AI-enhanced version of the model's face, body, or voice. That consent must spell out the scope, purpose, duration, and pay for the replica's use; routine photo retouching is excluded; and existing power-of-attorney agreements that had signed away digital-replica rights became void the day the law took effect. Violators face damages plus penalties of up to 100 percent of actual damages, or 300 percent if the violation was willful.
A companion measure, New York's Digital Replica Law, effective January 1, 2025, targets the broader "rights-grab" contracts the fashion and entertainment industries have historically used: an agreement letting a company generate a digital replica in place of work a model would otherwise have done in person is void if it doesn't specifically describe the intended use, or if it was signed without the model having access to a lawyer or union representative. New York followed up again in December 2025, when Governor Kathy Hochul signed the Synthetic Performer Act, which will require advertisements to conspicuously disclose the presence of an AI-generated performer once it takes effect in June 2026.
The concern driving all three laws is the same one voice actors raised in Hollywood: a single day of photography can now be stretched, with the right consent language buried in a contract, into a synthetic performer's unlimited future use, undercutting the very booking a model was hired, and paid, to do in the first place.
Insurance Claims Reviewers
Health insurance is one of the few industries here where the fight over AI decision-making has been driven as much by patients and doctors as by the reviewers whose jobs are on the line, but the resulting rules protect a very specific kind of human judgment: the licensed clinical reviewer's. California's Physicians Make Decisions Act (SB 1120), sponsored by the California Medical Association and signed by Governor Gavin Newsom in September 2024, requires that any denial, delay, or reduction of care based on "medical necessity" be made by a licensed physician or qualified clinician rather than an algorithm alone; it also sets hard deadlines for review, including five business days for standard requests and 72 hours for urgent ones, and puts the state's Department of Managed Health Care in charge of auditing insurers' denial rates. The law took effect January 1, 2025, amid national anger over insurance-claim denials following the killing of a UnitedHealthcare executive in New York that December.
UnitedHealthcare itself is the defendant in the highest-profile case testing these questions in court. Families of two deceased Medicare Advantage patients sued the company in 2023 over its use of nH Predict, an algorithm built by a UnitedHealth subsidiary that allegedly overrode treating physicians' recommendations and cut off coverage for skilled nursing care; the plaintiffs cite a roughly 90 percent reversal rate among denials that patients appealed. A federal magistrate judge ordered UnitedHealth to turn over broad internal records on the tool in March 2026, with a class-certification ruling expected later in the year.
Cigna faces a parallel set of suits over PXDX, an algorithm plaintiffs say let company doctors deny claims "in batches of hundreds or thousands" — one medical director reportedly denied 60,000 claims in a single month, spending, according to a widely cited ProPublica investigation, an average of 1.2 seconds per case. A federal court allowed core claims in that case to proceed in 2025, rejecting Cigna's argument that delegating medical-necessity decisions to the algorithm was a permissible exercise of its discretion under the plans it administers.
Content Moderators and AI Raters
This profession sits at an unusual intersection: these are workers whose labor makes AI systems usable in the first place, which means their fight isn't only about AI taking their jobs; it's about being asked to train the very systems that might replace them, sometimes while being squeezed on pay along the way. YouTube Music's content operations team in Austin, Texas, contract workers employed through Cognizant who correct AI-generated errors in the platform's music metadata, voted 41-0 to unionize with the Alphabet Workers Union-CWA in April 2023, after staging Google's first-ever labor strike that February over a retaliatory return-to-office mandate. The National Labor Relations Board ruled Google and Cognizant joint employers, and the unit ratified its first contract in 2024.
A more pointed version of the same dynamic played out in 2025 among Google's "super raters" — contractors, many with advanced degrees, hired through the Hitachi-owned firm GlobalLogic to evaluate and rewrite the Gemini chatbot's answers and Google's AI-generated search summaries for accuracy and tone. After months of organizing over pay disparities and job insecurity, more than 200 of them were abruptly laid off in August and September 2025; several had already filed unfair-labor-practice charges with the NLRB alleging retaliation. Internal documents reported by Wired indicated Google was working to automate the rating process itself, feeding a fear among workers that, in the words of one, they were being used to "replace themselves."
The trauma side of the job has drawn its own organizing response. In mid-2025, the Global Trade Union Alliance for Content Moderators, backed by the UNI Global Union, published the first international set of safety protocols for the profession, calling for limits on daily exposure to graphic content, an end to unrealistic productivity quotas, sustained mental-health support, and an explicit right to unionize. The group's accompanying survey found that 81 percent of content moderators believe their employers do too little to protect their mental health, even as the disturbing material they review increasingly trains the very AI systems being built to moderate in their place.
Autonomous Hauling
Examining the rapid deployment of self-driving trucks and the legislative gaps in protecting human drivers' livelihoods and safety.
AI in Patient Care
Analysis of AI diagnostics and robotic assistance, focusing on the critical human-in-the-loop roles in patient empathy and complex decision-making.
Robotics on the Line
Investigating the impact of advanced robotics on factory floors and the need for new statutory definitions of human supervision and skill.
Automated Instruction
Reviewing the rise of AI-driven learning platforms and the irreplaceable role of human teachers in fostering critical thinking and social development.
Ready for Action?
Our research provides the foundation for real policy. Join us in building a future where human work is valued and protected.
Whether you organize on the shop floor or manage a small enterprise, your operational knowledge informs our legislative drafting across all five platform areas.