Public Procurement and Artificial Intelligence: The First Cases at TAR Lazio
The Regional Administrative Court (TAR) of Lazio, in its recent ruling no. 4546 of 3 March 2025, addressed a case related to the use of artificial intelligence in public procurement from two different perspectives.
The first concerns the substantive aspect of a bid that includes, as a service provision method, the significant use of artificial intelligence resources alongside the technical part. The second concerns the use of the same artificial intelligence tools by the defenders of the party whose bid was deemed less competitive, to challenge the legality of the awarding of the first-ranked bids.
The Case
The case originates from an appeal filed by a company participating in a public service contract (cleaning and sanitation), to challenge the award to the winners who had prospected, among other things, the use of artificial intelligence in their bid.
With specific reference to the use of artificial intelligence, the appellant argued: a) the technical impossibility of implementing the use of artificial intelligence as envisaged by the winning bidder in its offer; b) the awarding of excessively high scores to the latter precisely in relation to the use of artificial intelligence envisaged in the winning bid.
In support and in proof of its arguments, the applicant claimed to have approached the same artificial intelligence platform mentioned in the successful bid, asking for clarifications precisely on its use, claiming that the answers provided by the said AI system were not compatible with what was stated in the successful bidder’s proposal.
The Position of TAR
The TAR found the appellant’s arguments to be unfounded, noting how, contrary to what was suggested in the appeal, the winning company had not envisaged a generic or undefined use of AI, but rather a targeted and specific use, limited to “mathematical/statistical support and data processing tools to improve the efficiency and quality of the service”.
The TAR also deemed the appellant’s contestation to be superficial and without merit, observing that the scoring criteria used by the evaluation committee were far more complex and could not be reduced solely to the AI component.
Furthermore, the TAR questioned the probative value of the appellant’s arguments, emphasizing that objections to the use of AI must be based on concrete evidence rather than mere hypotheses or simple inquiries.
Finally, the TAR reaffirmed that judicial review of the evaluation committee’s decisions is only permitted in cases of manifest unreasonableness, illogicality, or error in the committee’s operations. In the absence of such issues, the discretionary assessment of the committee is beyond judicial review.
Conclusions
In this ruling, the TAR acknowledges the widespread adoption and general use of artificial intelligence tools, correctly emphasizing the importance of a thorough and context-specific evaluation of their application, both in relation to the established criteria for the tender and to the specific applications proposed by the participants.
Although the ruling reaffirms well-established principles regarding the discretionary powers of evaluation committees, it nonetheless provides an opportunity for reflection.
Technological innovation must indeed be welcomed within public procurement and judicial contexts, while also recognizing its limitations and maintaining the fundamental principles that govern administrative decision-making and related legal proceedings.
Regardless of the specific circumstances of this case, the increasing reliance on artificial intelligence in public procurement suggests the need to develop guidelines or best practices that ensure transparency, uniformity, and, consequently, comparability in the assessment of proposals involving these technologies.
