Geneva School of Social Sciences Geneva School of Social Sciences

Research Center for Corruption Studies

The Research Center for Corruption Studies (RCCS), directed by Professor Emanuela Ceva at the School of Social Sciences of the University of Geneva and by Dr Maria Paola Ferretti , approaches the cross-disciplinary study of corruption as a deficit in the accountability chain linking officeholders to their mandates. It explores such questions as:

To what extent could different conceptualizations of corruption shape measurment instruments, policy-making and public perceptions?
Should we seek a unified understanding of corruption to make sense of misuses of power across cultural contexts, institutional set ups, and public as well as private organizations?
What are the duties of public servants to sustain institutional functioning against corruption?
What accountability practices can fight corruption and enhance public trust in institutions?

Funded on the ERC/SNSF Advanced Grant “The Margins of Corruption” the RCCS promotes rigorous, conceptually grounded, and practice-oriented research, training, and outreach activities to enhance the understanding of corruption and anticorruption in democratic public institutions and beyond.

Executive Board

RCCS Newsletter

RCCS Annual Reports

Should an AI Minister Fight Corruption? A View from Office Accountability

Emanuela Ceva, Geneva RCCS

05 May 2026

Albania's appointment of Diella, an AI-generated “minister” associated with public procurement, has made global headlines. The official government profile presents Diella as Minister of State for Artificial Intelligence, while Reuters reported that Prime Minister Edi Rama introduced it as a way to make public tenders “100% free of corruption.” The promise is politically powerful: remove human discretion from procurement, and corruption will disappear. The promise is also too quick.

The question is not mainly empirical regarding whether AI can help anticorruption. It can. AI systems can detect anomalies, compare bids, flag suspicious patterns, and make procurement files more transparent. The harder question is ethical and concerns whether AI could and should replace officeholders’ direct engagement, characterized by reflective and critical albeit fallible exercise of public power. From the perspective developed at the Geneva Research Center for Corruption Studies (RCCS), the answer is a qualified “no.”

The RCCS studies corruption as a deficit in the accountability chain linking officeholders to their mandates. In Political Corruption: The Internal Enemy of Public Institutions, Maria Paola Ferretti and I argued that corruption is not only bribery or personal enrichment. It is an unaccountable use of the power of office: a use for an agenda whose rationale is unjustifiable by reference to the power mandate. This matters because corrupt institutional action can occur even when no one pockets money.

Seen from this perspective, Diella's “ministerial” framing is ethically risky. Public procurement is not a mechanical exercise. It requires judgment: officeholders must interpret eligibility criteria, assess trade-offs in case of disagreement, and generally act in ways aligned with their best and bona fide interpretation of the raison d’être of their institution. If the decisive choice is attributed to a system whose reasoning cannot be reconstructed into reasons that officeholders can own, institutional authorship becomes blurred. Officeholders may then say that “the computer decided.” That is not anticorruption. It is responsibility deflection.

This concern does not entail a rejection of AI. In Automating Anticorruption?, María Carolina Jiménez and I argued that machine-learning tools may support anticorruption only if they strengthen, rather than weaken, office accountability. Diella could therefore be designed as a sentinel, not as an adjudicator. It may flag risk indicators, identify unusual bidding patterns, generate structured comparisons, and help oversight bodies focus attention. But final decisions should remain with identifiable officeholders who exchange reasons, answer objections, and can be reviewed by other officeholders and by the public, thus claiming co-authorship of any institutional act.

Anticorruption is not achieved by evacuating human discretion from institutions. It is achieved by structuring discretion so that those who exercise public power can give and demand reasons for how that power is used. Only under those conditions AI can uphold institutional functioning as an accountable work in progress.