The MCC-AI-High-Risk document offers a full legal toolkit for public buyers acquiring AI systems classified as high-risk under the AI Act. It transforms regulatory language into procurement-ready obligations and includes optional add-ons to reflect the diversity of public sector needs. Here’s how it breaks down:
A. Definitions
Establishes a shared vocabulary—crucial in AI contracting. Includes clear definitions for concepts like “Reasonably Foreseeable Misuse,” “Substantial Modification,” and the treatment of different types of Data Sets.
B. Essential Requirements
These are direct translations of AI Act requirements into contractual obligations, including:
- Risk management systems with continuous review and documented testing.
- Data governance, including bias assessment, provenance tracking, and representativeness of datasets.
- Technical documentation & instructions for use as deliverables.
- Logging and traceability through automatic recordkeeping.
- Transparency obligations to support interpretability and oversight.
- Human oversight mechanisms, including interfaces and override procedures.
- Robustness, accuracy, and cybersecurity aligned with Annex H measures.
C. Supplier Duties
Mandates quality management systems, conformity assessment, cooperation with rights impact assessments, and mandatory corrective action pathways.
D. Data Use Rights
Specifies ownership, usage rights, destruction duties, and potential handover of both public and third-party data sets—with flexibility through Annex B.
E. AI Register & Audit
Public bodies can request and disclose system information for AI registries. Audit rights are broad, and non-compliance can trigger remedy obligations. Suppliers bear compliance costs unless in full conformity.
F. Costs
Defaults to no additional fees unless otherwise agreed, reinforcing fiscal discipline.
Annexes A–H
Provide detailed templates for describing the AI system, data handling, technical documentation, accuracy metrics, and governance safeguards.
💡 Why it matters?
Procurement shapes the AI ecosystem. These clauses are a first-of-its-kind contract toolkit designed to give public authorities control—not just over what they buy, but how it’s built, monitored, and adapted.
This document does three things at once:
- Turns regulatory obligations into enforceable terms
- Offers ready-made language to avoid reinventing wheels for every tender
- Sets a de facto benchmark for “what good looks like” in responsible AI procurement
The MCC-AI-High-Risk is how the AI Act begins to take real-world effect. By hardwiring requirements for risk management, explainability, bias mitigation, and data traceability into the procurement process, public buyers can finally say: compliance isn’t optional—it’s contractual.
What’s Missing?
While the clauses are comprehensive, there are some areas that still need refinement or clarification:
- Balance of obligations: Optional clauses on explanations (Art. 14) and AI register disclosures (Art. 19) might become essential in practice. There’s a risk of fragmented uptake without clearer guidance from the Commission.
- SME-specific considerations: The clauses are heavy for smaller vendors. There’s no dedicated adjustment mechanism for simplifying obligations based on supplier size or capacity.
- Integration with national frameworks: These are pan-EU clauses, but AI procurement is often local. There’s no mention of how they map onto national contract law or sector-specific procurement rules.
- Handling of open-source models: While the clauses mention open-source under AI Act Article 53(2), there’s little operational guidance on how to contractually enforce obligations when the underlying model is freely available.
Best For:
- Procurement officers drafting tenders or contracts for AI-enabled systems in education, health, justice, etc.
- Legal teams within government bodies building frameworks aligned with the AI Act.
- AI suppliers navigating compliance while competing for public contracts.
- Regulatory consultants and AI ethics teams advising public institutions on operationalizing the AI Act.
Source Details:
Model Contractual Clauses for the Public Procurement of High-Risk AI, European Commission, Community of Practice on the Procurement of AI, February 2025 Draft.
Document status:
Non-binding. Issued as part of a working group process within the Public Buyers Community on AI, coordinated by the European Commission. These clauses are expected to evolve alongside AI Act enforcement practice.
Legal context:
Anchored in Chapter III of the EU AI Act, especially Articles 9–15, and related transparency and accountability requirements. Builds on the definitions and principles laid out in the AI Act but remains adaptable to specific contract terms via annex customization.
