Independent Consultant

Conceptualizing the need for the effective policy solutions such as sectoral standards and indicators to counter corruption during development and deployment of AI tools to prevent corruption.

• Interoperability of governance approaches
• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI

This submission is grounded in practical experience gained during my tenure as Chairwoman of the Corruption Prevention Commission of the Republic of Armenia, including leading the development of the electronic platform for asset declarations. The challenges and risks identified reflect direct institutional practice, particularly in designing and overseeing algorithm-based systems within a governance framework. These insights were subsequently developed into a structured research proposal and publication, which further examine the potential risks of unregulated or insufficiently governed AI systems in corruption prevention. The social, economic, ethical and technical implications of AI are critical, as the distinction between digital and traditional corruption is increasingly blurred. As a result, an unstandardized and unharmonized AI tools, combined with unethical governance, may reinforce existing vulnerabilities and diminish political accountability rather than strengthen it. Interoperability of governance approaches is essential to ensure that AI systems can be effectively regulated across jurisdictions. The absence of harmonized standards and regulatory frameworks creates enforcement gaps and limits accountability, particularly in cross-border contexts. While different regions (e.g. EU, OECD partners) move toward harmonization, existing approaches remain fragmented and insufficiently aligned. Transparency, accountability, and human oversight are essential to ensure that AI algorithms do not enable or conceal corruption. The use of AI in governance, in particularly in corruption prevention, introduces significant risks related to bias, lack of oversight, and insufficient legal safeguards. These risks are further exacerbated by the absence of concrete standards and indicators to assess algorithmic decision-making systems, as well as limited mechanisms to hold both public authorities and private developers accountable. As a result, existing frameworks often fail to adequately regulate non-human decision systems, leaving individuals exposed to unaccountable and potentially harmful outcomes. These priorities together underline the need for developing sector-specific standards and measurable indicators to ensure that AI serves as a tool against corruption, rather than enabling it.

One of an important cross-cutting is the absence of sector-specific integrity standards for the development and deployment of AI in corruption prevention. Across OECD countries, ethical rules and codes of conduct play a central role in shaping public sector behavior and ensuring accountability. However, these integrity frameworks have not been adequately extended to the design, development, and use of AI systems, in particular those applied to corruption prevention, detection, and investigation. Moreover, integrity standards for the private sector actors developing such systems remain limited or non-existent. This creates significant risks. The problem of bias in machine learning models directly affects the identification of corruption risks. For instance, in asset disclosure systems, algorithms may be trained in ways that "correct" discrepancies between income and expenditures, effectively preventing the triggering of red flags. In such cases, the manipulation or misuse of AI is not formally recognized or regulated as a form of corruption due to the absence of clear legal definitions, standards, and accountability mechanisms. At both national and international levels, there is a regulatory gap: AI systems used in governance often lack adequate testing protocols, oversight mechanisms, and enforceable responsibilities for both public authorities and private developers. As a result, potentially harmful or manipulative algorithmic practices can remain undetected and unaddressed. To address this gap, traditional corruption risk standards should be adapted to the AI context. This includes the development of sector-specific standards and measurable indicators that can guide the design and deployment of AI tools, as well as enable effective monitoring, auditing, and accountability of algorithmic systems in practice.

The governance gaps in the selected thematic areas have direct and tangible effects on countries with evolving rule of law systems, including Armenia and the broader Eastern Partnership region. In practice, the introduction of AI tools into governance, in particular in areas such as corruption prevention, has outpaced the development of adequate regulatory frameworks. While digitalization efforts, including asset declaration systems and data-driven oversight tools, aim to enhance transparency, the absence of clear standards on algorithmic accountability, interoperability, and human oversight creates new vulnerabilities. These include risks of biased decision-making, limited exploitability, and insufficient mechanisms to challenge or audit automated outputs. A key challenge is the lack of sector-specific standards and indicators to assess the functioning and integrity of AI systems. Existing legal frameworks often do not sufficiently regulate non-human decision-making processes, leaving gaps in responsibility and enforcement. This is particularly relevant where public authorities rely on automated systems to detect inconsistencies or corruption risks, but lack the tools to verify how these systems operate in practice.

The AI Dialogue can play a critical role in articulating the need to bridge existing international governance frameworks with the emerging risks associated with AI, including those related to corruption. While instruments such as the OECD principles on AI and the United Nations Convention against Corruption (UNCAC) provide important foundations, they were developed without fully accounting for the implications of AI-driven systems in governance. As a result, there is a growing gap between established anti-corruption frameworks and the realities of digital corruption, including the manipulation or strategic design of algorithmic systems to obscure accountability. The AI Dialogue offers a platform to explicitly recognize this gap and to advance a shared understanding that existing frameworks require adaptation and recalibration. This includes integrating AI-related risks into anti-corruption standards, developing common terminology, such as the recognition of misuse of AI as a form of corruption and identifying minimum requirements for transparency, accountability, and human oversight in algorithmic systems. Furthermore, the Dialogue can help highlight the need for greater coherence and interoperability across jurisdictions, particularly in cross-border contexts where AI system and private sector actors operate beyond single regulatory regimes. By convening governments, international organizations, private sector actors, and civil society, the AI Dialogue can also catalyze discussions on the development of sector-specific standards and indicators, ensuring that AI tools used in governance contribute to integrity and accountability rather than creating new vulnerabilities.

The AI Dialogue should build upon existing international anti-corruption and integrity frameworks, while explicitly integrating them with AI governance developments. Key instruments include the United Nations Convention against Corruption (UNCAC), the OECD Anti-Bribery Convention, the OECD Public Integrity Framework and Indicators, the Council of Europe Criminal and Civil Law Conventions on Corruption (with GRECO monitoring), and initiatives such as the Open Government Partnership (OGP) and programs led by the World Bank and UNODC. However, these frameworks were developed largely in a pre-AI environment and do not adequately address the risks of political corruption in digital spaces, particularly those enabled by algorithmic systems. Today, AI-driven content curation, targeting, and amplification mechanisms used by social media platforms can influence political discourse, shape public opinion, and potentially be manipulated to conceal undue influence or corrupt practices. A key gap is the lack of enforceable requirements governing the design and deployment of algorithms used in politically sensitive contexts. While anti-corruption frameworks focus on public officials and institutions, they do not extend sufficiently to private sector actors developing and operating these systems. The AI Dialogue can add value by articulating the need to merge and operationalize anti-corruption and AI governance standards, ensuring that core principles, such as transparency, accountability, integrity, and oversight, apply equally to algorithmic systems. This includes developing shared standards and indicators for AI used in political and governance contexts, and addressing the role of private platforms in preventing corruption risks.

Evidence based report, which includes package of recommendations with the model sectoral standards which can guide governments but also the tech companies in developing AI tools for corruption prevention, as well as indicators against which the AI tools can be measured by the civil society or any third-party auditor.

Partner with the following organizations, OECD, UNODC, World bank, OGP and government initiatives on information technologies.

Standards, requirements and indicators of transparency and accountability for the algorithmic decision-making systems used to prevent corruption at policy and legislation levels

As an example to be used is sectoral regulations and standards developed to encourage the responsible design and deployment of artificial intelligence tools, such as those involving the processing of judicial decisions and data. The CEPEJ European Ethical Charter on the use of artificial intelligence (AI) in judicial systems and their environment is one example striving for the accountability of the AI in judiciary.