Big Story: What International Research Reveals About Managing AI Risk in Government

Key Takeaways

  • Government technology leaders are facing a fast-moving set of questions about AI governance, workforce roles, and the boundary between innovation and risk.

  • McKinsey's 2026 AI Trust Maturity Survey finds steady progress on responsible AI practices overall, but persistent gaps in strategy, governance, and oversight of agentic AI.

  • Security concerns remain the top barrier to scaling agentic AI, and confidence in how organizations respond to AI incidents has declined even as incident rates hold steady.

  • Looking at how AI governance is evolving internationally can give state and local leaders a fuller picture of where the risks and best practices are heading.

Across government, technology leaders are wrestling with a similar set of questions. How should agencies weigh innovation against risk as AI tools take on more responsibility? How should staff roles change as efficiency gains reshape day-to-day work? And which organizations are managing this transition well enough to be worth studying closely?

These questions reflect a broader shift taking place across the public sector. AI is becoming embedded in core processes, from how agencies handle casework to how they detect security threats, and that shift is forcing leaders to rethink governance, staffing and risk management.

Global Government Forum study examines what it takes for governments to build durable digital capability in this environment. The report identifies a set of foundations that underpin digital success across government organizations, ranging from workforce visibility to public trust in digital services.

A central theme in the findings is the relationship between government and the private sector. Agencies need enough internal technical, commercial, and delivery expertise to act as an informed customer, rather than depending entirely on vendors to define solutions. It also calls for clear frameworks to guide decisions about which capabilities should be built in-house, and which are better handled through outside partners.

Governments need a clear, system-wide view of their suppliers, contracts and dependencies. Without that visibility, agencies risk fragmented technology environments and reduced market leverage. The recommendation is for governments to engage proactively with vendors to shape priorities and encourage competition.

McKinsey's 2026 AI Trust Maturity Survey examines how organizations manage trust as AI systems gain greater autonomy. The survey frames responsible AI practices as a core requirement for safely scaling the technology, since the consequences of failure grow as systems move from making suggestions to taking independent action.

Overall maturity in responsible AI practices continues to improve, but certain areas are lagging. Strategy, governance and controls for agentic AI remain underdeveloped, with only about 30% of organizations reaching higher maturity levels in these dimensions. Maturity also varies widely by region and sector, with some industries and parts of the world moving faster than others.

Perhaps most notable for government leaders, the survey finds that security and risk concerns are the leading barrier to scaling agentic AI, and that inaccuracy and cybersecurity remain the most frequently cited risks as adoption grows. At the same time, confidence in how organizations respond to AI-related incidents has declined, even though the frequency of incidents has stayed roughly the same.

Taken together, these findings point to a consistent message for government leaders. Building the internal capability to evaluate technology, manage vendors, and govern AI responsibly is becoming as important as the technology itself. Agencies that treat trust, governance, and workforce readiness as ongoing priorities will be better positioned to manage the risks that come with deeper AI adoption in the years ahead.

Build the right capabilities for what’s coming. Book a quick call with the Fractional Source team.

Quick Hit News:

  • Pottsgrove School District in eastern Pennsylvania approved a contract with the company Creatively Focused to build a closed-source AI system for its Pupil Services Department, which oversees special education programs. The system will operate within the district's own documents and procedures rather than the open internet, handling repetitive administrative tasks, delivering monthly insights to staff, and aiming to free up administrators for more direct work with students, at an annual cost of $10,000.

  • Oregon's new electric vehicle road fees, taking effect next year, will require most EV owners to pay either a per-mile charge of about 2 cents or a flat fee of $340 every two years, expected to bring in roughly $67 million for the State Highway Fund during the 2027 to 2029 budget cycle. Medium-duty corporate delivery fleets, including more than 11,000 electric vans owned by Amazon, were left out of the new law, a gap lawmakers say they intend to address in the 2027 session.  

  • California and Texas confirmed they are part of a new Anthropic cyber defense initiative that offers state, local, tribal, and territorial governments up to $15 million in credits to use AI tools such as Claude Security and Claude Code to find and fix system vulnerabilities. California's Department of Technology said a security engineer using the tool surfaced 40 vulnerabilities in a fraction of the normal time, while Texas Cyber Command said the partnership will help its team build the agility needed to keep pace with emerging threats.

  • The Las Virgenes Municipal Water District in Calabasas, California, completed a pilot project with OceanWell that lowered a desalination filtration pod into one of its reservoirs to test technology designed to purify seawater using ocean pressure. The company estimates it could cut energy use by 40% compared with traditional desalination plants. Officials plan to lower a pod 1,300 feet into the ocean this fall for weeks of testing, a step toward a yearlong demonstration project that could eventually supply 20-25% of the district's water from the sea.

  • Washington, D.C. launched a beta version of its redesigned DC Govt. website, its first major update in more than ten years, built on the open-source platform Drupal 11 with two-factor authentication, full site translation through Google Translate and an integrated events calendar. The site meets Web Content Accessibility Guidelines standards and will soon add an AI-powered search feature.

For the Commute:

The One About the Future of Benefits Delivery (The Government Huddle)

Chris McDermott, head of Social Security solutions at LexisNexis Risk Solutions and a former federal government leader, discusses how modernization, customer experience, identity verification, and public trust intersect in the delivery of government benefits. The episode focuses on how agencies can improve customer experience while still protecting programs from fraud, improper payments, and identity risk. McDermott’s perspective is useful because he brings both public-sector operating experience and private-sector identity expertise to the issue. The takeaway is that modernization cannot just mean faster digital tools. Benefits systems need stronger identity verification, cleaner data, and better fraud controls, but they also need to preserve access, dignity, and empathy for the people who depend on government services.

Resources & Events:

📅 Collaborative Leadership Leads to Housing for All (Virtual - June 24, 2026)

This webinar is on approaches to advancing attainable housing solutions in communities across the country. The session will examine how local leaders build consensus, foster collaboration among stakeholders, and develop practical housing strategies that meet residents' needs. Panelists will discuss how a collaborative, solutions-oriented approach to housing can support meaningful progress and help strengthen communities. Details →

📅 NACo Annual Conference and Exposition (New Orleans, Louisiana - July 17-20, 2026)

This event brings together elected and appointed officials from the nation's 3,069 counties, parishes and boroughs for 4 days of workshops, committee meetings, general sessions and policy discussions. The event includes NACo's ten policy steering committees and the Annual Business Meeting, where membership elects new leadership and sets federal advocacy priorities for the coming year, as well as opportunities to connect with peers and explore solutions from corporate partners. Details →

📊 Report Spotlight: Digital Government Outlook 2026 (OECD)

The OECD report finds that governments have largely laid the foundations for digital transformation but are struggling to translate them into improvements for citizens. Drawing on the Digital Government Index and OURdata Index across 36 member and 8 accession-candidate countries, the report says that to move from digital ambition to actual public-sector performance amid tight budgets, rapid AI advances, and fragile public trust, governments must set five priorities: stronger data governance for reuse, wider uptake of digital public infrastructure, modernized investment and procurement, robust trust frameworks for AI, and more proactive, human-centered service design. Read →

Insight of the Week:

Public pension systems were built around a model of continuous full-time employment with a single employer. An analysis from the Reason Foundation finds that traditional defined benefit pensions reward long uninterrupted careers and penalize part-time work, job changes, and career breaks, patterns common among women balancing caregiving responsibilities, leaving most female employees without meaningful benefits if they leave before vesting. For government leaders facing workforce shortages in roles largely held by women, reviewing vesting schedules and exploring portable benefit options could help recruit and reduce turnover among employees.

Was this email forwarded to you? Subscribe here to get Fractional Leader delivered to your inbox every Tuesday & Thursday.