The Paradox of AI-Driven Anti-Corruption in the Public Sector: New Corruption Risks in AI Procurement

In recent years, public organisations have rapidly expanded their engagement with AI to innovate their anti-corruption efforts. AI is being piloted and used to tackle different forms of corruption through various functions and across different government areas, from healthcare and welfare to justice and elections (Gerli & Mattoni, forthcoming). While AI-based Anti-Corruption Technologies (AI-based ACTs) can reduce discretion and promote transparency and accountability (Köbis et al., 2022), they also introduce new risks (Odilla, 2024; Köbis & Starke, 2022). As public authorities often lack in-house technical expertise, they frequently rely on public procurement to engage with private-sector providers (Gerli & Odilla, forthcoming). Nevertheless, these technologies’ procurement procedures remain opaque, also shaped by legal and technical challenges undermining AI procurement in the public sector more broadly (McBride et al., 2024) and underdeveloped policy frameworks. This paper aims to delve deeper into the opacity of public procurement for AI-based ACTs. The paper draws on in-depth interviews conducted across diverse European countries – Italy, Germany, Estonia, and Cyprus –, desk review of official and unofficial documents, and elements of participant observation. Findings suggest that AI-based ACTs can generate new corruption risks throughout the whole procurement cycle. Each phase – pre-tendering, tendering, and post-tendering – exhibits distinct configurations of monopoly, discretion, and accountability, thereby creating conditions for different corruption opportunities to emerge. These results underscore the need to update anti-corruption and procurement regulations to address AI-specific challenges, close exploitable grey areas, and balance procurement flexibility with accountability. Through its contribution, the paper offers a critical assessment of AI-based ACTs in the public sector and raises awareness of the risks associated with their adoption.