← Governance Research Brief

Governance Research Brief · Edition 3 · 18 September 2026

Published every Friday

The Governance Research Brief selects consequential new research, official reports and institutional analysis on sovereign artificial intelligence (AI), public administration and public judgement.

Each edition separates what the evidence establishes from the questions it raises for government, then considers the implications for Kenya, the Global South and the wider world without treating those settings as interchangeable.

Government accountability: risk, public trust and procurement

The Blind Spots Inside Public Authority

In this week's edition

Five new and recent publications show how the risks governments measure, the institutions they trust, the funding they can trace and the markets they shape determine what public authority can actually govern.

Ahmed Nadhir Omar

Published · 18-minute read · Five research readings

Selection rationale

Five publications, chosen for consequence rather than volume

Four selections were published between 12 and 18 September 2026. The fifth, published on 8 September, sits just beyond the one-week window and is included because its treatment of procurement and market-shaping power bears directly on the edition's argument. None appeared in an earlier edition.

5consequential publications
4published 12–18 September
1published 8 September
The governance test

Can the institution see the problem, identify who holds authority, act on the evidence and explain what changed?

Selected publication sources

Five institutional blind spots, considered together

Artificial intelligence and sector vulnerabilityMapping U.S. Federal AI Governance Against Sector VulnerabilityHo Ting Hung, Angelica Chowdhury, James Teague, Simon Mylius, Spencer Michaels, Peter Slattery, Alexander K. Saeri and Neil Thompson · 14 September 2026Institutional trust as administrative evidenceUpdated OECD Guidelines on Measuring Trust and Trust Drivers in Public InstitutionsOrganisation for Economic Co-operation and Development (OECD) · 16 September 2026Democratic resilience and conflict riskThe Global State of Democracy 2026: Democracy in an Age of ConflictInternational Institute for Democracy and Electoral Assistance (International IDEA) · 15 September 2026Development finance outside public viewMaking Development Co-operation More Effective: Progress Report 2026Organisation for Economic Co-operation and Development (OECD) and United Nations Development Programme (UNDP) · 15 September 2026Procurement as market-shaping powerPublic Procurement in the National Interest: Reflections from the CMAUnited Kingdom Competition and Markets Authority · 8 September 2026

Artificial intelligence and sector vulnerability

Mapping U.S. Federal AI Governance Against Sector Vulnerability

Ho Ting Hung, Angelica Chowdhury, James Teague, Simon Mylius, Spencer Michaels, Peter Slattery, Alexander K. Saeri and Neil Thompson · 14 September 2026 · Research preprint

What the research says. The authors examine 684 United States federal documents that address artificial intelligence, coding their coverage of 14 sectors and 24 risk areas before comparing the results with vulnerability judgements from a three-round Delphi study involving 272 experts. Public administration, national security, information and scientific services receive comparatively broad attention. Finance and health care receive less, even though experts judge them highly vulnerable. Finance received the highest vulnerability rating in 17 of the 24 risk areas but appeared in fewer than 30 per cent of documents covering any particular risk domain. Across the corpus, security, robustness and governance receive more attention than social, economic, environmental and multi-agent risks, while most mentions do not develop into detailed governance measures.

Limit of the evidence. The evidence describes attention within AI-specific federal documents rather than the full body of law. It excludes generally applicable legislation, state and local measures and many sector rules. The classifications rely partly on large language models, and the method does not weight documents by legal significance. The paper identifies where closer investigation is warranted; it does not prove that every apparent gap is an absence of governance.

Why this matters for public institutions

KenyaA national AI framework can appear comprehensive while leaving the services with the greatest public consequences thinly governed. Kenya needs a sector-by-risk map covering health, revenue, financial services, social protection, policing, land and county licensing. Information and Communication Technology Authority standards, sector regulators, the Office of the Data Protection Commissioner, national departments and counties should be able to see where controls overlap and where no institution has accepted responsibility.
Global SouthGovernments adapting models designed in better-resourced jurisdictions may copy their distribution of attention as well as their rules. Domestic vulnerability may sit elsewhere because service structures, informality, infrastructure, language and enforcement capacity differ. A credible map must include ordinary law, sector regulation and subnational practice.
Wider implicationsRegulatory activity should not be mistaken for regulatory coverage. Governments need to compare what policy documents discuss with where informed observers expect harm, then weight instruments by authority, test whether requirements can be implemented and trace responsibility across institutions.

The full preprint sets out the document-coding method, Delphi study and sector-by-risk findings behind this assessment.

Open Original Source

Institutional trust as administrative evidence

Updated OECD Guidelines on Measuring Trust and Trust Drivers in Public Institutions

Organisation for Economic Co-operation and Development (OECD) · 16 September 2026 · Official methodological guidelines

What the research says. The OECD moves the measurement of institutional trust towards more formal official statistics. Its proposed core questionnaire covers trust in national and local government, parliament, police, courts and political parties, alongside citizens’ expectations of reliability, responsiveness, openness, integrity and fairness. It also asks whether public agencies would use personal data legitimately, whether governments would regulate new technologies responsibly and whether people are satisfied with identification, birth registration and benefits administration.

Limit of the evidence. The guidelines warn that trust scores can conceal different phenomena. Critical mistrust can support democratic accountability, whereas entrenched distrust may foster disengagement. Cross-country comparisons require caution because response styles, institutional meanings, survey modes and sampling methods differ. This is a measurement standard, not new evidence about current trust levels.

Why this matters for public institutions

KenyaA Kenyan module developed with the Kenya National Bureau of Statistics could distinguish national government, county government and specific frontline institutions, while testing questions in Kiswahili and other languages before national use. County-level samples would make the evidence relevant to devolved services. Findings should lead to named changes in procedure, staffing, communication or review, rather than another popularity score.
Global SouthComparable measures can reduce dependence on occasional international perception indices. Their value depends on local adaptation, representative sampling and inclusion of people who are difficult to reach through digital or telephone surveys. Imported wording may otherwise erase the difference between formal authority and the institutions citizens actually encounter.
Wider implicationsTrust can inform public management without becoming a target to maximise. Independent collection, transparent methods and publication of uncertainty protect the measure from political manipulation and allow governments to compare citizens’ expectations with administrative performance.

The OECD publication provides the complete measurement framework, core questionnaire and methodological guidance.

Open Original Source

Democratic resilience and conflict risk

The Global State of Democracy 2026: Democracy in an Age of Conflict

International Institute for Democracy and Electoral Assistance (International IDEA) · 15 September 2026 · Global assessment

What the research says. The report draws on 155 indicators from 23 data sources, producing nearly two million observations across 174 countries from 1975 to 2025. In the five years to 2025, 98 countries, representing 57 per cent of those assessed, declined in at least one factor of democratic performance, while 55 countries, or 32 per cent, improved in at least one. Freedom of expression recorded the widest deterioration. Rule of law remained the weakest principal category, with 71 countries classed as low-performing and 29 recording category-level decline.

Limit of the evidence. The report notes that 2025 had 65 state-based conflicts, the highest number in the Uppsala Conflict Data Program series since 1946. Its more careful finding is not that democratic decline mechanically causes war. Conflict is concentrated in places already affected by instability, and democratic breakdown and violence may arise from the same crisis.

Why this matters for public institutions

KenyaThe report directs attention beyond elections to the institutions that make political competition governable: courts, legislatures, predictable enforcement, civic space and local participation. Formal participation cannot carry legitimacy if citizens cannot see how their representations alter budgets, plans or services. County assemblies, participation records and administrative review are part of democratic consolidation.
Global SouthRecurring elections are insufficient evidence of democratic resilience where rights, legislative oversight and enforcement remain uneven. External support that concentrates on electoral events while neglecting courts, public administration, local government and civic organisations may leave the institutions required for peaceful contestation unfinished.
Wider implicationsDemocratic quality is a security concern as well as a constitutional one. The global series shows widespread stagnation alongside concentrated decline, while cautioning against a simple causal claim. Rule-of-law institutions and legitimate channels for disagreement belong within long-term peace and security policy.

The full International IDEA assessment contains the global data, indicator definitions, country evidence and methodology.

Open Original Source

Development finance outside public view

Making Development Co-operation More Effective: Progress Report 2026

Organisation for Economic Co-operation and Development (OECD) and United Nations Development Programme (UNDP) · 15 September 2026 · Multilateral monitoring report

What the research says. The 2023–26 monitoring round draws on reporting by 44 partner-country governments and 93 development partners, supplemented by civil society organisations in 37 countries, trade unions in 16 and private-sector representatives in 20. It finds that 62 per cent of the 52 countries and bilateral development partners assessed for the relevant Sustainable Development Goal indicator reported a decline in multi-stakeholder monitoring frameworks. Development partners make limited and declining use of partner-country data and systems, while medium-term funding predictability has stagnated.

Limit of the evidence. The share of development co-operation recorded on national budgets fell from 61 per cent in 2018 to 41 per cent in the latest round. Only three countries met at least four of the five elements used to define a quality mutual-accountability mechanism. The evidence comes from a voluntary process with a limited and changing country sample and should guide country dialogue rather than serve as a league table.

Why this matters for public institutions

KenyaExternal finance that sits outside national and county budget systems weakens Parliament, county assemblies, auditors and citizens at the same time. Kenya needs a single view of development resources linked to the programme-based budget, the Integrated Financial Management Information System and county plans, with forward commitments disclosed early enough to affect allocation decisions. Projects implemented through counties should state who funds recurrent costs, who owns data and assets, and what happens when external support ends.
Global SouthCountries cannot exercise ownership over finance they cannot predict or see. Fragmented projects, new financiers and private-sector arrangements can outrun accountability structures designed for traditional aid. Development partners must provide usable forward plans and accept domestic reporting arrangements rather than create parallel ones.
Wider implicationsDevelopment effectiveness depends on more than aggregate financial flows. The distribution of information determines who can plan, scrutinise and learn. As financing becomes more diverse and geopolitical, mutual accountability must include newer actors without allowing complexity to become an excuse for opacity.

The joint OECD and UNDP report provides the monitoring results, country evidence, indicators and supporting annexes.

Open Original Source

Procurement as market-shaping power

Public Procurement in the National Interest: Reflections from the CMA

United Kingdom Competition and Markets Authority · 8 September 2026 · Official policy report · Selected from the preceding 30 days

What the research says. The Competition and Markets Authority argues that public procurement should shape competitive, innovative and resilient markets rather than operate as a sequence of isolated purchases. Its recommendations call for clearer system leadership, informed choices about what to aggregate and what to devolve, procurement data that reveal market structure and outcomes, lower barriers to entry and an innovation framework for frontier technologies. It also urges systematic collection of machine-readable bid data to detect collusion.

Limit of the evidence. An accompanying estimate suggests that, if bid rigging affected 2 per cent of United Kingdom procurement, taxpayers could overpay by about £1 billion each year, although that estimate rests on assumed prevalence rather than direct national measurement. The report reflects United Kingdom markets, law and institutional capacity; its recommendations travel more readily than its numerical estimates.

Why this matters for public institutions

KenyaNational ministries, state corporations and counties often buy from the same technology, construction, health and professional-services markets without a shared view of supplier concentration or performance. Common data standards, market analysis and framework terms can coexist with county discretion. For artificial intelligence and digital infrastructure, procurement should preserve audit access, interoperability, staff knowledge, service continuity and the ability to change suppliers.
Global SouthPublic purchasing is one of the few policy instruments large enough to influence domestic markets directly. It can widen entry and build local capability, but poorly designed localisation or complex bidding rules may protect incumbents. Shared regional assessment, standard clauses and pooled market intelligence can improve bargaining power without forcing every institution into the same contract.
Wider implicationsProcurement determines which firms grow, which technical standards prevail and whether public bodies retain operational control. Strategic purchasing requires clear trade-offs, usable data and officials with authority to manage long-term public value.

The official report presents the Competition and Markets Authority’s recommendations, supporting analysis and evidence on procurement markets.

Open Original Source

Synthesis

Public authority depends upon an unbroken chain of visibility and action

Read together, the publications show where an institution can lose sight of a public problem, the authority responsible for it or the power required to change course.

VulnerabilityWhat is exposed?
MeasurementWhat becomes visible?
AuthorityWho may decide?
ResourcesWhat enters public systems?
Market powerWhat does public purchasing shape?

Three questions for institutional practice

  1. Where does governmental attention diverge from public vulnerability? A useful artificial-intelligence or digital-governance review should compare the risks named in policy with the sectors, populations and administrative processes most exposed in practice, including county and local systems.
  2. Which information must enter public systems before accountability becomes possible? Trust evidence, development finance, bid data, model logs and contract performance all need a responsible custodian and a route into budgets, oversight and correction.
  3. Can public institutions change course when the evidence changes? Measurement has little value if officials cannot revise a procurement, alter a service process, challenge a development partner, strengthen a safeguard or explain the decision to citizens.

Public authority becomes credible when institutions can identify their exposure, locate responsibility, retain the capacity to act and learn from the consequences. That obligation applies as much to an artificial-intelligence system as it does to a county service, a development programme, a democratic institution or a public contract.

The five sources