Artificial Intelligence in Public Procurement 2026

Artificial intelligence in public procurement: procurement of AI systems, EU AI Act, awarding AI development services and using AI in tendering processes.

Definition: Artificial intelligence (AI) in the public procurement context comprises both the procurement of AI systems and services by public contracting authorities and the use of AI tools to support the procurement process itself – both are subject to specific legal requirements, in particular under the EU AI Act (2024).

Last updated: January 2026 · Legal basis: EU AI Act (Regulation 2024/1689); GWB/VgV; BVergG 2018; Directive 2014/24/EU

AI procurement as a growing topic

Public procurement of AI systems and services is rapidly gaining importance, as authorities, municipalities and public institutions increasingly use AI-based solutions for administrative workflows, decision support, forecasting and citizen services. At the same time, procuring AI poses particular challenges for contracting authorities: AI systems are often difficult to specify, their quality depends on training data and algorithms, and there are significant risks for fundamental rights and discrimination.

EU AI Act and procurement

The EU AI Act (Regulation 2024/1689, in force from August 2024) classifies AI systems by risk class and imposes different requirements on providers and deployers – requirements that directly affect public contracting authorities as deployers of AI systems.

Risk classes under the EU AI Act:

Prohibited AI (Art. 5): Systems that are banned (e.g. social scoring by authorities, biometric mass surveillance) – may not be procured or deployed.
High-risk AI (Art. 6 in conjunction with Annex III): E.g. AI in law enforcement, education, employment, critical infrastructure – subject to strict transparency, documentation and oversight obligations; must be CE-marked.
Limited-risk AI: Transparency obligations (e.g. chatbots must disclose that they are AI).
Minimal risk: No specific requirements.

Public contracting authorities procuring high-risk AI systems must ensure that the systems comply with the EU AI Act requirements; this must be included as a performance requirement in the procurement documents.

Public procurement specifics for AI procurement

Procuring AI services confronts contracting authorities with typical procurement problems: drafting specifications is difficult, evaluating algorithmic quality, data protection, explainability and avoiding vendor lock-in.

Recommendations for AI tendering procedures:

Functional specifications: Description of the desired outcomes and quality parameters rather than technical specifications of the algorithm.
Explainability as an award criterion: For high-risk AI, the explainability of the system's decisions should be evaluated.
Non-discrimination: Evidence that the AI system does not produce impermissible discrimination.
Data protection impact assessment: Obligation to carry out a DPIA for high-risk data processing.
Interoperability and portability: Open interfaces to minimise dependency on the provider.

AI in the procurement process itself

Beyond the procurement of AI systems, AI is also increasingly being used to support the procurement process itself – for example for the automated review of bids, the detection of bid-rigging cartels or to support bid evaluation.

The use of AI in the procurement process raises its own legal questions:

May an AI fully take over the evaluation of bids?
What human oversight is required for automated award decisions?
How can transparency and documentation obligations be met for AI-supported decisions?

Under the EU AI Act, AI systems used in the award of public contracts qualify as high-risk AI (Annex III no. 5) where they are used for decision-making – they are therefore subject to the corresponding requirements regarding transparency, human oversight and documentation.

Public funding and AI procurement

The federal and state governments have launched various funding programmes that support the public procurement of AI solutions. The Federal Ministry of Digital and Transport (BMDV) and the federal AI strategy provide that public bodies should act as lead markets for AI solutions.

FAQ

May public contracting authorities use AI systems for administrative decisions? Yes, but subject to the requirements of the EU AI Act, in particular for high-risk AI: human oversight, transparency, documentation and conformity assessment are mandatory.

What must be considered when purchasing AI-as-a-Service (cloud AI)? In addition to procurement law, data protection (GDPR), the EU AI Act, remote-access and third-country rules, and requirements arising from the NIS2 Directive must be observed.

Can a tender simply require "AI" as a requirement without further specification? No, specifications must be concrete enough for bidders to submit a complete bid; "AI" as a general requirement is too unspecific and would breach the principle of transparency.

Last updated: January 2026 All information provided without guarantee. For legally binding advice, please consult a law firm specialising in public procurement law.