NetEase Yidun launches AI-native security system with Agent Guard
Overview
NetEase Yidun, the content and risk-control security unit of NetEase's enterprise services business, launched an AI-native security product system at the 2026 World Internet Conference Light of Internet Expo in Wuzhen on November 3.
Its Agent Guard checks whether an AI agent's actions match the user's intent, enforces permissions, flags risky tool calls and blocks behavior, while guardrails screen model inputs and outputs at the agent layer.
The company says its CMA moderation agent cut average complaint-ticket review time from 35.7 minutes to 12.3 minutes at 99.2% accuracy for one customer handling more than 2.5 million tickets a month. It also says its DeepGuard face anti-spoofing scored 99% recall and 96% precision against screen re-capture attacks in phase testing. These figures come from the company's own statements.
Written by AI from the articles below · updated Nov 4, 2:37 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- PandailyNewsNetEase Yidun launches Agent Guard to police what AI agents do
AINetEase Yidun launched an AI-native security product system at the 2026 World Internet Conference Light of Internet Expo in Wuzhen, with Agent Guard checking whether an agent's actions match user intent and blocking risky tool calls. The company said its CMA moderation agent cut average complaint-ticket review time from 35.7 minutes to 12.3 minutes at 99.2% accuracy for one customer handling more than 2.5 million tickets a month. Its DeepGuard face anti-spoofing scored 99% recall and 96% precision against screen re-capture attacks in phase testing.
Heat Trend
Not enough continuous observations to show a trend yet.