US unions push AI guardrails as most agencies adopt generative AI

Union wins across several states

Public-sector unions in the United States are negotiating new protections against workplace AI, according to The Decoder. The push comes as an Ernst and Young survey indicates that close to 70 percent of US government agencies have already adopted generative AI tools.

The reported agreements vary in form. In Oakland, a strike vote in October led IFPTE Local 21 to secure retraining programmes in place of AI-related redundancies. In Pennsylvania, Governor Josh Shapiro assured SEIU Local 668 that people would oversee decisions during a trial of ChatGPT Enterprise from OpenAI.

Elsewhere, the Transport Workers Union in Ohio gained the right to veto the use of autonomous vehicles. These deals show unions seeking a direct say in how automation is introduced, rather than resisting it outright.

New state laws limit surveillance and job cuts

Legislation is moving alongside collective bargaining. California Governor Gavin Newsom signed a ban in September on AI monitoring software, commonly described as bossware, and on dismissals carried out automatically by algorithms.

New York has gone further for its own workforce. Since late 2025, state law has shielded public employees from being displaced by AI. Together, the two states set a template that other legislatures may follow as AI spreads through government work.

Why unions see urgency now

Employers are adopting automation to reduce costs, according to Kung Feng, a labour researcher at UC Berkeley. That pressure makes workers vulnerable to both constant digital monitoring and decisions made without human review.

Union leaders argue the answer is not to block the technology. Liz Shuler of the AFL-CIO, Lorena Gonzalez, and John Samuelsen of the Transport Workers Union have said they back AI systems that help workers, the Guardian reports. Their condition is that a person must retain control over every AI application used at work.

What the trend means for public workers

The combination of high adoption rates and new bargaining wins suggests AI in government is becoming a standard subject of negotiation. Workers in agencies using generative AI may increasingly see contract clauses on oversight, retraining and monitoring.

For now, the reported measures remain uneven. Some protections come from state law, others from individual agreements, and their reach beyond the states involved is unclear. Readers should treat the survey figure and individual deals as reported by The Decoder and the Guardian rather than independently confirmed.