MarkAny (CEO Choi Choi) and Nflux (CEO Lee Min) have been granted two ‘Artificial Intelligence (AI) Reliability Certifications (CAT)’ by the Ministry of Science and ICT (Minister Lee Jong-ho) and the Korea Telecommunication Technology Association (TTA, Chairman Son Seung-hyeon). It was announced on the sixth that it had been received.

The AI ​​reliability certification system is aimed toward private AI operators and developers who want to voluntarily secure AI reliability and is predicated on AI reliability group standards. TTA, a personal certification agency, analyzes the danger aspects of services and products using AI technology and evaluates and awards business operators compliance with requirements to make sure reliability.

Since 2021, the Ministry of Science and ICT and TTA have been developing and distributing 'Reliable AI Development Guide' for every field and establishing technical requirements to secure AI reliability. Last December, based on the contents of the event guide, it became the primary in Korea to secure international compatibility with the international standards related to AI reliability: ISO/IEC TR 24028 (reliability overview), ISO/IEC 23894 (risk management), and ISO/IEC 22989 (terminology). The AI ​​Reliability Information and Communications Organization standard ‘Requirements for Enhancing AI System Reliability’ was also established.

MarkAny acquired this certification for its ‘AI image search and object movement path tracking solution.’ We identified AI model errors equivalent to accuracy errors in missing person recognition and bias toward specific races, genders, and ages as risk aspects, verified AI model bias removal, established AI system reliability test plans, and established corporate governance to secure AI reliability.

Nflux's 'AI convergence mine detection module v1.0', a wise control specialist, is a system that reads ground penetrating radar (GPR) images and provides high-level evaluation functions equivalent to determining whether a mine is a landmine and the kind of landmine. We identified mine detection accuracy errors and the shortage of explainability of mine detection results as risk aspects, and focused on verifying the appropriateness of the information construction method and securing the explainability of the AI ​​model's judgment results.

TTA, a personal certification agency, provided support for improving the reliability of AI services and products by providing consulting services to business operators regarding supplementary needs throughout the means of conducting AI reliability verification.

This certification is the primary case because the introduction of the private autonomous AI reliability certification system last October. It’s meaningful in that it has established a personal autonomous system to make sure the reliability of domestic AI services and products.

The Ministry of Science and ICT said, “In a situation where efforts to secure AI trust and safety are fierce all over the world, we expect that a certification system through an expert organization will probably be a great model that may promote safety, trust, and innovation at the identical time.” “We plan to strengthen the security foundation of the domestic AI industry by upgrading the AI ​​reliability certification system to reflect the rapidly developing generative AI technology and market changes and continuing to expand AI reliability certification best practices,” he said.

Reporter Jang Se-min semim99@aitimes.com