Selectstar said on Wednesday its in-house red-teaming technology, Startiming, was officially adopted by ACL 2026, a leading natural language processing conference. The paper addresses an automated red-teaming method to verify the safety of large language models by attempting harmful prompts to find vulnerabilities. In tests on 17 models, Startiming achieved a 74.5 percent average attack success rate, it said, exceeding AutoDAN-Turbo by 13.5 percentage points.

The government is changing how it supports AI-related vouchers, reducing direct aid for small firms to buy and process data and shifting toward integrated support for corporate AI transformation (AX). The 2026 data voucher budget is set at 9.35 billion won, down 66.6 percent from a year earlier, amid declining revenue for the Information and Communications Promotion Fund. A new AX one-stop voucher is budgeted at 30 billion won, but its higher threshold limits access for smaller firms.