Although AI is becoming critical infrastructure for states, almost half of governments still lack mandatory regulatory mechanisms to oversee it. This is one of the main conclusions of the OECD Digital Government Outlook 2026, recently published by the Organisation for Economic Co-operation and Development (OECD), which analyzes the progress of digital transformation in 36 member countries and 8 countries seeking accession.
The report reveals a significant fracture in how governments are approaching AI governance. While 53% of countries have opted for binding regulatory frameworks that require algorithmic transparency, auditability and formal accountability mechanisms, 40% continue to rely solely on voluntary guidelines and ethical principles with no legal force.
Institutional risks
The OECD warns that this situation could become one of the main institutional risks of the next decade. The growing automation of administrative processes, the use of algorithms to support public decisions and the expansion of generative artificial intelligence systems in public administration require much more robust control structures than those currently in place.
The document also identifies the near-universal absence of audits after AI systems are deployed. Although governments usually assess technology investments beforehand, they rarely analyze years later whether the initial promises of efficiency, cost savings or service improvements have actually been met.
The OECD identifies 200 AI use cases for governments
Beyond these warnings, the OECD Digital Government Outlook 2026 confirms a profound shift in how the digital transformation of the state is understood. The digital maturity of administrations is no longer measured by the simple digitization of procedures or the existence of online portals, but by the capacity to incorporate disruptive technologies at the core of their operations and, especially, by the institutional strength capable of governing them.
Critical infrastructure
The digital maturity of states, according to the report, is no longer measured by the mere migration of procedures to web portals, but by their capacity to absorb disruptive technologies at the core of their operational functioning. In this new scenario, AI, both predictive and generative, has become the central axis of transformation strategies.
Through the Digital Government Index, the OECD finds that AI is no longer an experimental pilot project, but a critical infrastructure tool used in detecting tax fraud, optimizing healthcare resources and automating citizen services.
However, the report warns that the speed of technology adoption risks outpacing the capacity of states to guarantee an ethical, safe and transparent deployment. With an overall DGI average of 0.70, the report shows that true international leadership is determined by the strength of the institutional governance underpinning these algorithms.
Digital maturity of countries
The OECD methodological framework assesses the digital maturity of countries through six key dimensions. Analyzing the incorporation of artificial intelligence from this perspective makes it possible to identify both the progress achieved and the main weaknesses that still persist in public administrations.
The first dimension is Digital by Design, which holds that AI should not be incorporated as a later technological add-on, but should be part of the very design of public policies and government services from their conception.
The second is Data-Driven Public Sector. Since AI systems depend directly on the information with which they are trained and operate, the quality, availability and interoperability of public data become a determining factor for their effectiveness.
The third dimension, Government as a Platform, highlights that the efficient deployment of AI requires shared infrastructure within the state, including computing capabilities, common systems and language models that can be used by different public agencies.
The citizen experience
The Open by Default dimension, for its part, has historically recorded one of the lowest scores in the index, with an average of 0.59. In the field of AI, this weakness translates into limited transparency about the algorithms used to support or make decisions that directly affect citizens.
The fifth dimension, User-Driven, focuses on the citizen experience. From this perspective, tools such as virtual assistants and AI-based conversational agents should facilitate interaction with the administration and simplify access to public services.
Finally, the Proactiveness dimension represents the most advanced level of digital transformation. In this scenario, AI is capable of anticipating citizens' needs and acting before they submit a request, for example by granting aid, renewing benefits or activating certain services automatically.
AI regulation
The most relevant policy finding of the Outlook 2026 lies in the profound global fracture over how to regulate AI in public administration. The report reveals a clear division in the oversight strategies implemented by governments:
53% of the countries surveyed have opted for a mandatory regulatory approach, establishing formal and binding legal requirements. These legal frameworks require algorithmic transparency, system auditability and explicit accountability mechanisms to mitigate discriminatory bias.
40% of states, by contrast, rely exclusively on voluntary guidelines, non-binding codes of conduct or abstract ethical principles with no force of law. The OECD issues a severe warning about this lagging 40%.
The absence of coercive regulatory tools and the lack of mandatory instruments for ethical impact self-assessment compromise institutional integrity. Without a strict legal framework, the automation of administrative decision-making risks violating fundamental citizen rights and opens the door to procedural opacity.
Cost-benefit analysis
One of the report's most critical diagnoses points to the budgetary governance of technology investments. While member states have pre-established methodologies for approving financial allocations for the acquisition or development of AI systems, the investment lifecycle shows an alarming structural gap after deployment.
The Outlook 2026 denounces the widespread omission of ex post cost-benefit analysis and long-term value proposition assessments.
AI-based technologies are introduced under the theoretical promise of generating operational efficiencies and substantial budget savings. However, public administrations practically never re-audit these systems years after implementation to compare projected performance metrics with the real costs of maintenance, the energy consumption of the models or the real impact perceived by society.
This financial blind spot prevents identifying the true return on investment and perpetuates inefficient spending on redundant or obsolete technology platforms.
Conclusions and recommendations
The OECD Digital Government Outlook 2026 conclusively demonstrates that AI has ceased to be a technology option and has become a central component of contemporary governance infrastructure. To prevent automation from eroding democratic trust and weakening the quality of spending, the OECD outlines three urgent regulatory and institutional mandates for states:
- Move toward binding legal frameworks: move beyond the stage of voluntary ethical recommendations and codify strict obligations of transparency and algorithmic auditability in public administration.
- Institutionalize ex post financial auditing: implement mandatory evaluation methodologies that scrutinize the real performance, operating costs and social impact of deployed AI systems.
- Enable data ecosystems for innovation: strengthen the open data reuse pillar of the OURdata index, facilitating secure technical access to public datasets to boost the development of AI models with high social value.
Only by immediately closing these control and governance gaps will governments manage to turn the inherent risks of AI into a dividend of efficiency, equity and public value for citizens, the report states.