AI Governance in Government: A New Framework for Public Sector Accountability

AI governance in the public sector is gaining ground and deservedly so.

getty

A newly published book provides a comprehensive framework for artificial intelligence governance within the public sector, addressing a critical need as government agencies accelerate adoption of the technology. The volume, Governing With AI: How the Public Sector Can Use Artificial Intelligence to Improve Performance, co-authored by Mark Fagan and Ben Gillies and published in February 2026, delivers fundamental principles and practical insights for public servants navigating this complex landscape.

While AI governance in the commercial sphere primarily focuses on regulating organizations that build or deploy AI for business objectives, public-sector governance carries a distinct mandate: governing the exercise of governmental authority through AI while preserving democratic values and the rule of law. This distinction necessitates a tailored approach that accounts for the unique constraints and responsibilities of state institutions.

AI Governance In The Public Sector

Public sector entities frequently face intense pressure to deploy AI rapidly. Personnel often experiment with generative AI and large language models (LLMs) in personal contexts, creating an internal perception that the organization is falling behind. However, launching isolated AI projects without a mature governance structure typically creates more risk than value.

A working definition helps frame the discussion: Public-sector AI governance encompasses the principles, structures, practices, and stewardship through which an entity ensures that its development or procurement of AI systems—and the deployment, use, and eventual retirement of those systems—are undertaken in a properly documented and effective manner while safeguarding democratic values, individual rights, public trust, and the rule of law.

This definition underscores the necessity of managing the full system development life cycle (SDLC) and the broader AI portfolio lifecycle, rather than treating governance as a one-time compliance checkpoint.

AI Is Likely Already In Their Midst

A common blind spot is the assumption that an agency is “AI-free” simply because no formal procurement approval has been granted. In reality, AI often enters organizations through two subtle vectors.

First, modern software-as-a-service (SaaS) platforms frequently embed AI capabilities—such as automated summarization or predictive analytics—deep within their feature sets. Vendors may not explicitly flag these components during procurement, leaving the agency unaware that AI is processing sensitive data.

Second, “shadow AI” usage occurs when staff leverage publicly available tools like ChatGPT, Claude, or Gemini via personal devices to draft memos, analyze data, or code. While seemingly innocuous, this practice introduces data privacy, accuracy, and accountability risks that fall outside official oversight.

Identifying this existing landscape requires an inventory approach, not a witch hunt. The goal is to surface ad hoc usage and integrate it into a structured governance framework that aligns with organizational policy and legal obligations.

A Definitive Guide For Public Leaders

Governing With AI arrives as a rare practical resource tailored specifically for government practitioners. At approximately 200 pages, the text serves as both a strategic playbook and a potential academic textbook, complete with discussion questions, essay prompts, and assessment tools.

Appendix D provides a self-assessment instrument comprising roughly forty overarching AI governance questions, designed for pre- and post-training evaluation. While the authors suggest a three-point rating scale, the framework is adaptable to various pedagogical preferences.

The publication builds on the authors’ established expertise; Fagan’s work at the Harvard Kennedy School’s Taubman Center for State and Local Government continues to produce influential research on AI risk management and specialized public-sector policy.

Public Sector Distinctiveness

The book identifies six structural challenges that differentiate public-sector AI adoption from the private sector:

• (1) Rule-based environment.
• (2) Limited resources.
• (3) Shifting priorities.
• (4) Lack of expertise.
• (5) Politics.
• (6) Complexity of government.

These factors compound one another. Unlike commercial enterprises—which can often allocate innovation budgets and recruit specialized talent quickly—public agencies operate under rigid procurement rules, hiring freezes, and legislative mandates. Legacy technology stacks further impede the integration of modern AI systems. The text offers strategies, tactics, and case studies to navigate these constraints across eleven chapters.

Risk And Reward In High-Stakes Environments

The stakes of public-sector AI errors are fundamentally different from commercial missteps. An algorithmic failure in a retail recommendation engine results in a lost sale; a failure in a system determining disability benefits, parole eligibility, or disaster resource allocation can alter the trajectory of a human life. Consequently, the standard for governance in government must exceed that of the private sector, yet historical attention to this domain has been insufficient.

Simultaneously, governance must not become so onerous that it paralyzes beneficial innovation. AI offers tangible public value: reducing administrative backlogs, detecting fraud, optimizing infrastructure maintenance, enhancing emergency response, and improving service delivery. An overly restrictive framework risks depriving communities of these advances.

The path forward requires balancing rigorous oversight with operational agility. AI governance must evolve into a mission-critical capability embedded in every adoption decision. As Johann Wolfgang von Goethe observed, “What is not started today is never finished tomorrow.” The work of building that foundation is underway.

Source link