Korean AI firms sign MOUs, win projects and NeurIPS spots
Flex, Konan Technology, ISTsoft, Xenon and Filer each announced a separate AI milestone on October 21, 2026, according to aitimes.com.
Five Korean AI companies announced separate developments on October 21, 2026, according to aitimes.com: Flex signed an enterprise AI memorandum of understanding with Microsoft Korea, Konan Technology agreed a defense AI partnership with Seoul National University of Science and Technology, ISTsoft was picked to build an in-house generative AI model for the Jeju Free International City Development Center (JDC), Xenon passed mass-production validation for vision AI inspection at a Nike factory in Indonesia, and Filer had two papers accepted to NeurIPS 2026.
Flex said the agreement with Microsoft Korea covers joint sales to Korean companies, product and technology integration, and a review of applying each other's solutions. Flex plans to link its organizational and talent data with Microsoft 365 Copilot to build enterprise AI that answers questions using the asker's job and organizational context.
Under its MOU, Konan Technology will develop AI-based defense solutions for intelligence, surveillance and reconnaissance (ISR), while the university handles faculty advising and joint research. The two will also pursue government projects together and link internships and practical training.
ISTsoft will build a work-support platform combining an sLLM and AI agents on internal servers separated from external networks. The project includes dedicated chatbots for areas such as development, duty-free business and budgeting, and a retrieval-augmented generation service drawing on about 3.8 terabytes of internal documents.
Xenon said its outsole inspection system, which uses high-resolution cameras and AI to detect surface defects, recorded a defect detection rate of up to 99.7% at the Indonesian plant. The company aims to put the mass-production model on production lines next year and to extend local work to shoe upper inspection and collaborative-robot process automation. Filer's accepted papers are CoAT, which helps AI read non-verbal cues such as vocal emotion and intonation, and GAPO, for stable learning from data containing errors; the company said the work will feed its brand safety and content moderation.
Quick answers
What did Flex and Microsoft Korea agree to do?
They signed an MOU covering joint sales to Korean companies, product and technology integration, and review of applying each other's solutions, including linking Flex's organizational and talent data with Microsoft 365 Copilot.
How accurate is Xenon's outsole inspection system?
Xenon reported a defect detection rate of up to 99.7% in mass-production validation at a Nike factory in Indonesia.
Which papers did Filer get into NeurIPS 2026?
Two papers: CoAT, on understanding non-verbal information such as vocal emotion and intonation, and GAPO, on stable learning from data containing errors.