Korea’s Public Procurement Service is developing an AI-specific evaluation framework for public procurement to enable the safe and effective adoption of AI products and services. The initiative introduces new assessment criteria covering AI performance, reliability, transparency, and risk management, addressing the limitations of traditional procurement standards. It aims to support trustworthy AI adoption while fostering innovation and market access for AI suppliers.
Initiative overview
Public procurement plays a strategic role in Korea’s AI ecosystem, representing a large-scale demand-side policy instrument. At the same time, Korea became the first country in the world to enact a comprehensive AI Basic Act, establishing overarching principles for trustworthy and responsible AI. However, despite this landmark legislation, clear operational definitions and classification criteria for AI-enabled products and services have remained underdeveloped in the procurement context. Existing public procurement evaluation systems—designed primarily for conventional goods and services—have therefore proven inadequate for assessing AI-enabled offerings, particularly with respect to algorithmic performance, data dependency, model reliability, and potential risks.
This initiative responds to that institutional and practical gap by developing an AI-specific evaluation framework tailored to public procurement. Rather than applying a single, uniform assessment model, the framework introduces procurement-instrument–specific AI evaluation approaches across major programmes, including innovation procurement, excellent product designation, and multi-supplier framework contracts. Each procurement scheme incorporates AI-relevant evaluation criteria aligned with its policy objectives and operational characteristics, recognising that AI products require different assessment logics depending on the purchasing mechanism. Instead of relying solely on legal definitions, the framework provides practical and evaluative criteria to identify and assess AI-enabled products and services within procurement processes. New assessment dimensions include AI model appropriateness, robustness, transparency, lifecycle risk management, and alignment with public sector use cases, complementing existing criteria on public value, performance, and innovation. This approach allows contracting authorities to more accurately distinguish genuinely AI-enabled solutions from conventional products, while ensuring that AI-specific risks are systematically considered.
Other relevant details
While still under development, the initiative has already informed the design of AI-specific evaluation tracks within Korea’s innovation and excellent procurement programmes. It is expected to improve procurement quality by enabling more accurate assessment of AI-enabled products and ensuring that genuinely high-quality AI solutions are adopted by public institutions. Clear evaluation criteria help prevent the entry of “AI-washed” products, reducing procurement risk and increasing trust among public buyers. In the medium term, the framework is expected to leverage public purchasing power to stimulate the AI market, support capable suppliers, and promote consistent and transparent AI procurement decisions across government.
Introducing AI into public procurement proved more complex than expected. Key challenges included rapidly evolving AI technologies, limited evaluation capacity within contracting authorities, and the lack of clear criteria to distinguish AI-enabled products from conventional solutions. Existing procurement evaluation systems were not designed to assess AI-specific characteristics, making a dedicated evaluation framework necessary. Despite a global AI boom, clear and operational standards for regulating and adopting AI in the public sector remain limited. Lessons learned highlight the importance of expert-based evaluation, adaptable criteria, and early engagement with technical experts. The initiative shows that AI procurement standards must address governance and lifecycle management, alongside performance, supported by continuous institutional capacity-building.
The initiative is designed as a flexible, non-binding guideline that can be embedded into procurement evaluation procedures and expert review panels. By translating abstract AI governance principles into concrete, procurement-ready evaluation standards, it reduces ambiguity for both public buyers and suppliers. Over time, the framework is expected to evolve through pilot applications, expert feedback, and alignment with emerging AI governance regulations under the AI Basic Act. It is designed to be scalable and replicable, offering a reference model for jurisdictions seeking to integrate trustworthy AI principles into public procurement systems where mature AI product classification regimes do not yet exist.
