AI Labor Tracker
SEIU Asked City College of San Francisco to Rule Out AI Layoffs. The College Struck Those Lines and Offered to Bargain Over AI's Effects Instead.
A marked-up bargaining proposal posted by City College of San Francisco shows where employers draw the line on AI. The college crossed out the union's job-security language but kept limits on AI monitoring and discipline, with exceptions. A Long Beach contract ratified by union members this week and a Toronto bargaining platform point the same way: unions are winning rules about how AI is used on workers, not promises about whether it replaces them.
32.9%
union membership rate of US public-sector workers in 2025, against 5.9% in the private sector (BLS, Feb 18, 2026)
At 10:48 a.m. on Oct. 29, SEIU Local 1021 handed City College of San Francisco a proposal titled “New Language: Artificial Intelligence (AI).” The union’s City College chapter officers work in offices like admissions, purchasing, student health and student conduct, according to the college’s labor relations page. Their contract with the college ran from July 1, 2023, to June 30, 2026, the same page shows, so the two sides are bargaining a new one.
The union’s proposal was direct. “Generative AI should not be used to enable the layoff of an employee or the eliminating of a position,” one line read. Another said employees “will not be laid off or downgraded as a result of implementation of AI systems.”
On Nov. 5, the college answered. It posted its counterproposal on its own website, marked up in the usual bargaining format, with a color key showing what each side added and deleted.
Every one of the union’s job-security sentences is struck through in the college’s color. What the college kept, and what it added, tells you more about where AI bargaining is heading than any national survey I’ve seen.
What the college crossed out, and what it kept
Here’s what the markup shows, according to the document’s own color key. I’ve read the strike-through lines against the key; this is a negotiating draft, and nothing in it is final.
The college struck the union’s opening sentence giving SEIU “the right to negotiate provisions to protect workers from involuntary job loss” as the college reshapes “tasks, workflows, or job classifications.” It struck a no-termination clause covering automation and “the combining of jobs or the elimination of jobs.” It struck the line saying technology “is not intended to replace the human element.” It struck both no-layoff sentences quoted above, and the union’s language on retraining for a comparable position or transferring workers when retraining isn’t feasible.
In their place, the college added a sentence of its own: “The District may determine the technology that will be used in connection with its operations to support student need and ensure efficiency.”
But the college didn’t strike everything. It kept, in edited form, a commitment that when it introduces new technology affecting matters within the scope of bargaining, it “shall negotiate the effects” of the changes with the union. It added that it would “ensure employees are properly trained on new technology.”
And it kept the heart of the union’s section on monitoring, with exceptions. The union proposed that “AI Systems shall not be used to” monitor, evaluate or discipline union members, or to collect data on individual performance, or to make automated decisions about employment status, assignments or working conditions.
The college left those words in. It added that “in compliance with applicable law, the District may use AI tools in the preparation of evaluation and disciplinary records” and performance records. And on automated decisions, it added five words that change the meaning: “without human review and approval.”
So under the college’s version, AI can’t make the decision about your job by itself. A person has to sign off. And the college keeps the right to decide whether AI does your work.
Who has a contract to bargain over AI: union membership, 2025
Percent of wage and salary workers.
How to read thisBars start at zero. Teal is the public sector, the finding. Gray is all workers, the private sector, and local government. 2025 figures are an 11-month average because October was not collected.
32.9%
public-sector union membership rate in 2025 (BLS, Feb 18, 2026)
Bar chart. Use the up and down arrow keys to move through 4 categories. A data table follows.
All workers: Union membership 10.0%
- BLS says the 2025 figures are 11-month averages. October 2025 was not collected during the federal shutdown.
Figure 1. Public-sector membership was 32.9%. The private-sector rate was 5.9%.
Source: U.S. Bureau of Labor Statistics, Union Members 2025, Feb 18, 2026. Chart: Post Labor Economics Download CSV Methodology
Data table
Local government, which includes city colleges, was 37.8%. The private-sector rate was 5.9%.
| item | Union membership |
|---|---|
| All workers | 10.0% |
| Private sector | 5.9% |
| Public sector | 32.9% |
| Local government | 37.8% |
The question
So here’s what I want to understand. Is City College’s line typical, and why do employers draw it there?
The answer matters because unions are the only private rulebook most American workers have for AI at work right now. Congress hasn’t written rules for AI at work, as Axios noted in July, and state rules are patchy. If you want to know what protections workers can actually win, contract language is the best evidence there is.
Where this bargaining happens
Start with who has a contract to bargain over.
The union membership rate for all US wage and salary workers was 10.0% in 2025, according to the Bureau of Labor Statistics. In the public sector, it was 32.9%, more than five times the private-sector rate of 5.9%. In local government, which includes city colleges and city halls, it was 37.8%. (The BLS notes that its 2025 figures are 11-month averages because October 2025 data weren’t collected during the federal shutdown.)
That’s why AI bargaining shows up first in places like City College. The office work AI is aimed at, scheduling, records, routine correspondence, student services, is done at public employers by people who often have a union. In private offices doing the same work, they mostly don’t.
And explicit AI language is still rare. The UC Berkeley Labor Center reviewed more than 500 collective bargaining agreements for its inventory of technology provisions, last updated in July 2025. It found relevant technology language in more than 175. “A few recent contracts explicitly reference artificial intelligence or generative AI,” the inventory says, “though most do not name AI or other emerging technologies directly.” Most rely on older clauses about “technological change” or “automation.”
It’s growing in some industries. Jon Schleuss, president of the NewsGuild-CWA, told Axios in July that his union had between 85 and 90 contracts with explicit AI provisions.
Some private-sector unions have won language closer to what SEIU asked for. Axios reported that a union at ZeniMax, the Microsoft-owned video game company, won provisions requiring AI to support workers rather than replace them, along with a requirement that management notify the union and negotiate before deploying new AI tools. At Politico, Axios reported, unionized reporters took management to arbitration over AI reporting tools introduced without the union’s agreement, and management took the tools down.
Notice what those wins have in common. The ZeniMax language is a statement of purpose plus a notice-and-bargain rule. The Politico case was enforced through a process the contract already had. Neither is a flat promise that no job will ever be cut because of AI.
Two more contracts this week
City College isn’t alone. Two other public-education and public-sector bargains in the news this week point the same way.
On Nov. 10, IAM Local 1930 announced that city workers in Long Beach, California, had ratified a tentative agreement. According to the union’s announcement, it includes 8.4% in general wage increases over three years and “new language to limit the use of Artificial Intelligence in discipline procedures.” The deal still needs City Council approval later this month, and the contract text isn’t public yet, so I can only report the union’s description.
The context there was a budget, not a chatbot. In July, the city manager proposed cutting 367 jobs, 158 of them vacant and 209 filled, mainly at libraries and public health. The union says the city agreed to reduce the workforce largely through unfilled vacancies and attrition. That’s a jobs protection won through ordinary bargaining. The AI piece of the deal is about discipline.
In Toronto, the Canadian Union of Public Employees Local 3902, which represents teaching assistants and other academic workers at the University of Toronto, began bargaining in October. Its platform, as The Varsity reported on Nov. 8, seeks protection against automation, layoffs and contracting out, and limits on AI used to monitor workers without consent.
The union’s statement to The Varsity is the closest thing to a worker’s voice in any of these documents. “Chatbots are being rolled out to replace actual TA work in multiple departments, including Mathematics and Engineering,” it said. The union said more than 900 members answered its bargaining survey, and two-thirds thought the university was on the “wrong track” with AI.
The university’s response was one sentence: it addresses labor matters “through negotiated collective agreements, and not through the news media.” That’s a reasonable policy. It also answers nothing.
Why employers draw the line there
There’s a first-principles reason the line sits where it does, and it isn’t stubbornness.
American labor law has long separated two kinds of bargaining. An employer generally has to bargain over the effects of many business decisions, such as severance, transfers and training, even when the decision itself belongs to management. The Supreme Court drew a version of that line for private employers in First National Maintenance Corp. v. NLRB in 1981. Public-sector rules vary by state, but City College’s existing contract already follows the same pattern. Its general provisions say the college “shall notify the Union and bargain the impacts/effects” of technology decisions made through its governance council, according to a July bargaining document posted by the college.
Seen that way, the college’s Nov. 5 markup is almost textbook. It refused to give up the decision: whether to use AI and whether that changes how many people it needs. It agreed to bargain the effects: training, process and protections around how AI is used on the people who stay.
There’s also a budget reason. A public college can’t promise that no position will ever be eliminated as a result of a technology, because it can’t promise what its funding will be. A no-layoff guarantee tied to AI would also be hard to enforce, since almost any future cut could be argued to have some AI connection.
The monitoring rules are different. They cost the employer little to grant, and they protect something workers care about a lot: not being judged by a machine. That’s why both City College and Long Beach moved on discipline and evaluation.
What workers are actually getting
My read: unions are winning real protections about how AI is used on workers and almost none about whether AI replaces them. The City College markup shows that line in a single page. The college kept limits on AI monitoring and discipline, added human review for automated decisions and promised training, and crossed out every sentence that would have stopped AI from being used to cut jobs.
That’s not nothing. A rule that no machine makes the final decision about your job, and that you can see and contest what’s used to evaluate you, is a meaningful protection. It’s also not the protection the union asked for first, and it isn’t the one most people worry about.
The honest limit is that this is one draft in an open negotiation. SEIU can push back, and the final contract may look different. Long Beach’s language isn’t public yet, and Toronto’s bargaining has just begun.
Back to City College
The college’s added sentence is the one I’d reread if I worked there: the District “may determine the technology that will be used in connection with its operations to support student need and ensure efficiency.” It’s a management-rights clause in one line. It’s also the reason the job-security paragraphs had to go, because both can’t be true at once.
The two sides are still at the table. The college’s own page shows the old contract expired June 30, and the AI article is one piece of a larger negotiation over pay and working conditions.
My recommendation, if you’re in a union or thinking about organizing one: when AI comes up at the table, ask for the effects you can win and enforce. That means notice before new tools arrive, real retraining with pay, human review of any automated decision about you, and the right to see the data used to evaluate you. Those are the terms employers are already conceding, and in the City College draft, they’re the parts that survived.