By late 2024, 82 percent of state chief information officer (CIO) organizations reported daily use of generative AI tools among employees, up from 53 percent the previous year, according to NASCIO. However, a Pew survey found only 24 percent of states had established data governance policies for generative AI, highlighting a rapid adoption outpacing oversight.
What happened
Several states—including New York, Pennsylvania, New Jersey, Colorado, and California—have launched generative AI programs for public employees recently, using different procurement methods and governance approaches. New York’s AI Pro, a generative AI assistant built by the state’s Office of Information Technology Services in partnership with Google Gemini, expanded in April 2026 from a pilot of 1,200 employees across eight agencies to over 100,000 state workers. While AI Pro is presented as a training tool, pilot users drafted memos, summarized documents, and processed agency work through the system.
Concerns center on transparency and data privacy: prompts entered by employees pass through Google’s infrastructure, and public documents do not clarify tenant isolation, data residency, or data usage limits. A 2025 audit by New York State Comptroller Thomas DiNapoli found the state’s AI policy insufficient in guiding agency use and called the gaps a “wake-up call.” Despite this, the rollout proceeded without resolving those issues.
New York published Appendix C-AI contract clauses restricting vendor training on state data and requiring audit rights, but it remains unclear if these protections apply in the Google partnership announced concurrently. The state also launched the FutureWorks Commission in March 2026 to recommend protections from AI-driven displacement, but AI Pro’s expansion preceded the commission’s report.
Other states took different paths. Pennsylvania established a comprehensive governance framework before deployment, signing an executive order in September 2023 to create a Generative AI Governing Board and codify principles. In March 2025, the administration negotiated a binding side letter with SEIU Local 668 to protect employees, including prohibiting AI use in disciplinary actions. By April 2026, ChatGPT Enterprise was deployed to 3,000 employees with mandatory training required before access.
New Jersey launched the NJ AI Assistant in July 2024 on Microsoft’s Azure OpenAI, later rebuilding it in March 2026 using AWS and Claude with LibreChat, an open-source interface. The state surveyed public-sector workers on AI before finalizing governance frameworks and tailored training by job function.
Colorado, after passing an early comprehensive AI law, replaced it with a narrower risk management framework effective January 2027. It deployed Gemini Advanced AI tools across 20 agencies with mandatory training and attestation. California expanded use of its “Poppy” AI assistant alongside AI sandboxes while developing vendor certification standards with a July 2026 deadline.
Why it matters
The rapidly growing use of generative AI in state governments presents data protection, transparency, and labor concerns. Deploying AI tools without established governance heightens risks around sensitive data, employee rights, and public accountability. Pennsylvania’s proactive governance and union engagement offer a model for minimizing these risks by building oversight before full rollout. Conversely, New York’s experience highlights the dangers of expanding AI use ahead of adequate policies.
Clear, public documentation of AI governance, mandatory employee training before use, and upfront labor-impact assessments are emerging best practices that states must adopt to ensure responsible AI implementation. As state governments increasingly employ AI, balancing innovation with protections is critical for trust and safety.
Background
Generative AI adoption in public sectors has accelerated since 2023 as states seek efficiency and improved workflows. National surveys show a sharp increase in AI usage by government workers, but governance frameworks lag. Several states have passed AI laws or executive orders requiring transparency, risk management, and labor protections. However, consistent application of these policies to state-deployed AI tools remains uneven, with challenges around vendor contracts, data residency, and employee engagement.
Sources
This article is based on reporting and publicly available information from the following source:
Read more AI Regulation stories on Goka World News.