During a 2025 IT forecast in January, Doug Robinson, executive director of the National Association of State Chief Information Officers, confirmed that generative artificial intelligence will be the most impactful emerging technology for state CIOs over the next several years.
NASCIO has recommended considerations for state CIOs developing AI roadmaps, and Robinson revisited three of those:
- Establish governance and oversight processes.
- Address data quality and sourcing.
- Expand AI workforce expertise and training.
In recent months, other associations have made recommendations that complement these considerations. As state governments expand their AI use cases, CIOs should weigh these best practices for AI governance.
Click the banner for more information about AI solutions for state and local governments.
Association Recommends Oversight Objectives
The American Association for the Advancement of Science once recommended best practices for AI governance among states, advising state governments to follow in the footsteps of the federal government.
AAAS promoted the concept of “trustworthy AI” and suggested steps to take for transparency in AI adoption. It called out three specific objectives:
- Mandate risk management practices. States should engage in AI impact assessments, monitoring, training and more to protect rights and safety.
- Require reporting and documentation. States can increase accountability by publishing an inventory of AI use cases and publicly reporting on their impact.
- Order careful acquisition policy. States should ensure effective AI procurement by testing technology, maintaining government control, promoting visibility and more.
Careful government oversight establishes boundaries for productive AI activity.
EXPLORE: Advanced processors are supporting the rise of invisible AI.
Experts Call for Strong Data Quality and Sourcing
The National Governors Association (NGA) and AAAS identify three major risks posed by generative AI: hallucinations, thwarted technical guardrails and bias. These risks speak to the need for AI to operate on quality data.
State CIOs should mitigate these risks through strong AI governance, the associations say:
- Hallucinations. Generative AI models don’t know what they don’t know. They can fabricate answers based on false data if left unchecked.
- Technical issues. Authorities can set guardrails against generating false alerts or narratives through generative AI, but it could bypass these guardrails potentially.
- Bias. Officials have long warned of skewed data that may harbor racial or other biases. NGA notes how a generative AI scenario depicted only white males when asked to show pictures of surgeons. This bad data could lead to flawed government actions.
Generative AI results are only as good as the data that feeds AI applications. If the data is poor, AI will be ineffective. Smart risk management paves the way for good data quality and sourcing.
SUBSCRIBE: Click the banner to get weekly updates from StateTech magazine.
States Leading the Way on AI Workforce Training
Several states have led the pack with their AI workforce development initiatives. The Public Sector HR Association last year lauded Pennsylvania for its ChatGPT Enterprise AI pilot. The commonwealth will build on lessons learned from the pilot to “inform the development of employee training and resource needs,” according to a government press release.
Meanwhile, the New Jersey Artificial Intelligence Task Force has four working groups, including Workforce Training, Jobs of the Future and Training Public Professionals. In the task force’s 2024 Report to the Governor on Artificial Intelligence, that working group asked New Jersey public sector employees about their knowledge and attitudes regarding AI.
The survey results showed that few state employees have used Generative AI, although most were optimistic about its potential:
- 88% said they knew “nothing at all” or “a little” about AI.
- 79% do not yet use generative AI tools at work.
- 73% expressed a desire to learn how to use generative AI at work.
- 67% believe generative AI will have more positive than negative impacts.
In 2024, the New Jersey Civil Service Commission introduced free training for the state workforce in collaboration with InnovateUS.
As the New Jersey AI report concludes, “One thing is certain: workers will need to train for change. Rolling out whole-of-workforce training in AI for the public workforce has set the stage for New Jersey to provide AI education for all.”
UP NEXT: AI isn’t new to cybersecurity, but some of its use cases are.
