Search results for AI Safety Report
AI & Enterprise
SKT\'s A.X K2 posts top-tier results in AI math reasoning evaluations
SK Telecom said its A.X K2, developed for a government-led sovereign foundation model project, posted strong results on math-reasoning benchmarks. The company said the model scored 29 out of 42 on the IMO 2026 evaluation, meeting this year’s gold medal cutoff, and tied for the top score on the MathArena AIME 2026 leaderboard with 97.1 percent accuracy. It also said A.X K2 achieved a perfect score in the KMO 2026 second round.
AI & Enterprise
K-AI models deployed across public administration, from R&D budget reviews to elderly care
South Korean AI models are being deployed across public administration, from government R&D budget reviews to services for seniors living alone. The Ministry of Science and ICT said domestic models from Upstage, LG AI Research and Naver are being integrated into central and local government work. The rollout includes a shared AI platform for government networks, development of science-focused models through the K-Moonshot programme and an AI competition using K-AI models.
AI & Enterprise
Big Tech AI alliances reshuffle as agent payments evolve
OpenAI has revised its partnership agreement with Microsoft, allowing it to offer its models on cloud services beyond Microsoft Azure, with Amazon Web Services moving to provide OpenAI models through its Bedrock platform. Some AWS customers are reported to be unenthusiastic, saying existing Bedrock models are sufficient. Competition is also intensifying over technologies for AI agent payments, including OKX’s Agent Payments Protocol and Stripe’s Link wallet. The article rounds up related moves by global and Korean firms.
