The Brief
Work & Economy 5 min read

94.1% Voted to Strike. Oakland's AI Rules Came Next.

NAVION

Share

When city workers in Oakland, California, voted on 22 October to authorize a strike, 94.1% said yes. The dispute was not mainly about pay. It was about artificial intelligence: whether software could be used to monitor, evaluate, discipline, or even replace public employees. Within days, the city and the union reached a tentative three-year agreement, and the strike planned for 5 November was put on hold.

The story is small in scale and large in implication. Public workers across the United States are starting to treat AI the way earlier generations of labor treated the assembly line: as something to be negotiated, not simply installed.

What Oakland’s workers actually won

The union at the center of this case is International Federation of Professional and Technical Engineers Local 21. Its members worry that AI systems could ramp up surveillance and score their performance inaccurately. Their demand was less a ban than a seat at the table.

“We just want the city administration to really be intentional about how it’s engaging with AI products,” said Julian Ware, an Oakland spatial data administrator and union member. His warning was practical: a shiny new tool can cost more, be harder to implement, and deliver worse results than the process it was meant to improve.

The tentative agreement includes protections such as a requirement that the city retrain or reassign workers displaced by AI instead of firing them. That detail matters. It does not stop the city from adopting AI. It changes what happens to people when it does, shifting the cost of automation from the worker to the employer.

A pattern spreading across public-sector unions

Oakland is one of many examples. According to a survey by Ernst and Young cited in the source, nearly 70% of state and local government agencies say they have adopted or are implementing generative AI. Adoption is moving fast, and workers are organizing at the same pace.

Pennsylvania offers a revealing case. The state became the first to run a pilot program with OpenAI, encouraging employees such as unemployment claim workers and social workers to test ChatGPT Enterprise for time savings. Steve Catanese, president of SEIU Local 668, said the pilot came with no guarantees or agreements for workers. “With generative AI, workers are building their own gallows sometimes,” he said, pointing to the risk that employees help train the very systems that justify cutting their roles.

The union then pushed Governor Josh Shapiro to agree to a policy guaranteeing that the state’s use of AI will include employee input and supervision, and that the technology won’t be used to discipline workers. Catanese expects the issue to return when the union negotiates a new contract in 2027.

Transportation is another front. Workers in states including Florida and New York are trying to block Waymo and city governments from deploying autonomous vehicles, calling them unsafe and a threat to jobs. They are hoping to repeat a 2024 result in Columbus, Ohio, where the Transport Workers Union won contracts giving it veto power over any Central Ohio Transit Authority plan to deploy autonomous vehicles. The union’s president, John Samuelsen, stressed that the TWU does not oppose all new technology. It supports systems that help with collision avoidance and improve bus and train timing.

Why “bossware” is the real story

Most headlines about AI and work focus on job loss. This story points somewhere subtler. Researchers use the term “bossware” for AI tools that employers use to monitor and supervise workers. According to an October report co-authored by Kung Feng of the University of California at Berkeley’s Labor Center, these tools can rank and evaluate workers for hiring, firing, and promotion. They can track location and activity, monitor productivity, scan biometrics for timekeeping, use cameras, wearables, and sensors, or apply analytics to predict emotions.

That is the part most coverage misses. A worker may keep a job and still lose control over it, if an opaque system decides how fast they should move or whether they deserve a promotion. Lorena Gonzalez, president of the California Federation of Labor Unions, argues that warehouse-style algorithmic pressure is now being brought into the public sector.

Her principle is a useful one for anyone thinking about AI at work: the technology can assist, but a human must remain in charge. She pointed to aviation, where computers have been able to fly a plane for years, yet every aircraft carries two pilots who can override the system with judgment and skill. “If we’re using a computer or AI, it has to be controlled by a human being, not the other way around,” she said.

What this means more broadly

There is a gap between how fast AI is being adopted and how fast rules are being written. Liz Shuler, president of the AFL-CIO, argues that policymakers have moved too slowly, so unions are using contract negotiations to build guardrails that lawmakers haven’t. Some legislators, especially at the state level, are making progress. The source notes that unions backed California’s No Robo Bosses bill, though the text cuts off before describing it fully.

The demands themselves are modest and concrete: advance notice before a new technology is introduced, information about how it works, and the right to bargain over its use. None of these block progress. They simply insist that the people affected by a tool have a say before it arrives. As Feng framed it, this is the contest over the future of technology in the workplace.

In short

  • Oakland’s city workers voted 94.1% to authorize a strike over AI, and the vote helped produce a tentative three-year deal with protections, including retraining or reassignment instead of firing.
  • Nearly 70% of state and local agencies report adopting or implementing generative AI, so these negotiations will keep multiplying.
  • Surveillance and evaluation tools, or “bossware,” may matter as much as outright job loss.
  • The core union demands are notice, information, and a right to negotiate, not a rejection of technology.
  • The guiding idea across cases is human oversight: AI can assist, but people stay in control.

Based on reporting from The Guardian.

Written by

NAVION