Editorial illustration for Unions secure AI guardrails as 70% of US agencies use generative AI
Unions Win AI Guardrails as US Agencies Adopt GenAI
Unions secure AI guardrails as 70% of US agencies use generative AI
Sixty-nine percent of US government agencies now use generative AI tools, according to a survey from Ernst and Young. That number is forcing a fight that public-sector unions are, for the moment, winning. Across the country, labor groups are extracting written commitments on retraining, oversight and veto power before agencies roll out chatbots and automated systems on their workforce.
The pattern looks similar in each case. A union identifies a specific AI deployment, whether it's a pilot program or a surveillance tool, and demands a human stay in the loop before management can act on the software's output. IFPTE Local 21 in Oakland got there after threatening to strike in October.
SEIU Local 668 in Pennsylvania negotiated directly with Governor Josh Shapiro's office over a ChatGPT Enterprise pilot. Ohio's Transport Workers Union went further, winning veto authority over autonomous vehicles entirely.
State governments are moving too. California banned so-called "bossware" surveillance tools and automated firing decisions in September. New York passed its own protections against AI-driven job loss late in 2025.
The common thread, labor leaders say, isn't blocking AI outright. It's making sure a person signs off before it changes someone's job.
US public-sector unions are winning more protections against AI surveillance and automation-driven job losses. The issue is growing more urgent as nearly 70 percent of US government agencies already use generative AI, according to an Ernst and Young survey.
Why this matters
For anyone building or selling AI tools to government, the Oakland and Pennsylvania cases are the template to watch, not the exception. A 70 percent adoption rate among agencies, per Ernst and Young, means procurement officers are now negotiating under contracts that specify human oversight, veto rights, and retraining clauses before a chatbot ever touches casework. That changes the sales pitch.
"Our model is more accurate" won't close a deal if Local 668's agreement with Shapiro's office already mandates a human review every ChatGPT Enterprise output in a pilot program. Developers who treat deployment as a technical rollout, rather than a labor negotiation, will keep getting blindsided by strike votes like IFPTE Local 21's in October. The Transport Workers Union's veto power in Ohio suggests this isn't isolated to white-collar unions either.
Expect contract language, not just agency policy, to start dictating where generative AI can actually run in the public sector. Watch whether these union wins get codified into state procurement rules, because that's the point where they stop being local labor stories and start being market constraints.
Related Reading
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google to End Free Access to Gemini AI for Non-Subscribers
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
From Archives ChatGPT adds new Wallet to let AI agents shop and check out for you
From Archives Claude’s Duet agent links Gemini meeting summaries, neurotic but helpful
Common Questions Answered
What percentage of US government agencies are currently using generative AI tools?
According to an Ernst and Young survey cited in the article, 69 percent of US government agencies now use generative AI tools. This widespread adoption has prompted public-sector unions to take action in securing protections for their members against potential job losses and surveillance risks.
What specific protections are US public-sector unions securing in AI deployment agreements?
Public-sector unions are extracting written commitments that include retraining programs, oversight mechanisms, and veto power before agencies roll out chatbots and automated systems. These protections address concerns about AI surveillance and automation-driven job losses across government operations.
How are the Oakland and Pennsylvania union cases influencing AI procurement in government?
The Oakland and Pennsylvania cases serve as templates for how government agencies must now negotiate AI contracts with procurement officers. These agreements specify requirements for human oversight, veto rights, and retraining clauses before any chatbot or automated system can be deployed, fundamentally changing how AI tools are sold to government entities.
Why is the 70 percent adoption rate of generative AI in US agencies significant for AI vendors?
The high adoption rate means that AI vendors can no longer rely solely on accuracy claims to close government deals, as procurement officers are now negotiating under contracts with strict guardrails. This shift in negotiating power represents a major change in the sales landscape for companies building or selling AI tools to government agencies.
Further Reading
- Nearly 70 percent of US agencies use generative AI, and unions are finally catching up with guardrails - The Decoder
- Survey Outlines Government's Progress on Modernization, AI - Government Technology
- Report: How to protect public sector workers against AI's rise in government - Route Fifty
- 2026 EY State and Local Government Modernization Survey Findings - Ernst & Young
- US public workers are fighting to protect their jobs from AI – and some are winning - The Guardian (syndicated)
