- Huawei claims its new Xinghe AI security stack can achieve ~95-99% detection or response rates for unknown threats and automated alarms through unified policy control.
- The update introduces more granular control at application level (micro-isolation), better endpoint defence for large model workflows, and tools to counter IoT and malware via “endpoint + network” defence.
What happened：Huawei rolls out Zero-Trust AI security across branch, campus and data centre
During Huawei’s Data Communication Summit at HUAWEI CONNECT 2025, the company unveiled the upgraded Xinghe AI Network Security Solution. It embeds a zero-trust framework aiming to cover enterprise branches, campuses, and data centres.
Some of the key features include:
- Zero-Trust Branch Access: The USG6000F branch security gateway combines centralized policy orchestration through the iMaster NCE-Campus controller, resulting in up to 99% automated threat response, with AI-based threat detection and an emulator engine to unpack unknown malware, claiming a 95% detection rate for unknown threats.
- Better asset recognition (over 95% accuracy), AI grouping, active scanning, and automatic creation of fine-grained isolation policies—particularly for east-west traffic, which refers to risks moving laterally across the campus network—are all features of the Zero-Trust Campus Interconnect.
- Zero-Trust Data Security: Dedicated to safeguarding big AI/ML models, the HiSec Endpoint detects kernel-level exploits during deployment, while an XH6655 intelligent computing firewall with antivirus engine checks for embedded malware during training. Additionally, features like defenses against prompt injection and one-click ransomware recovery.
Also read: Oracle brings AI to UK sovereign Cloud in $5bn Investment Push
Also read: Huawei builds SuperClusters to challenge global AI race
Why it’s important
The announcement is made at a time when businesses are more vulnerable to AI-powered threats because threat actors can automate attacks using large models and because businesses are implementing AI/ML workflows that could introduce vulnerabilities, such as in the government, healthcare, and meteorology sectors. Huawei’s action demonstrates how important zero-trust + AI integration is to security providers.
However, there are open issues:
- Real-world efficacy: The claimed 95% detection rates, automated responses etc., are impressive in lab or controlled conditions—but how do they perform under heavy load, in diverse environments, or when adversaries adapt?
- Complexity vs manageability: Zero-trust architectures and micro-segmentation introduce complexity. Enterprises may find managing fine-grained policies and ensuring they don’t hinder operations is challenging.
- Trust and supply chains: The article mentions large-model training with open-source frameworks may have supply chain vulnerabilities. While Huawei offers tools to detect “embedded malware,” verifying that these tools catch all threats is non-trivial.
All things considered, Huawei’s enhanced Xinghe platform promises improved endpoint defense, threat detection, and policy automation, advancing zero-trust thinking for the AI future. However, how well these features perform in situations where security and usability must be traded off, under actual threats, and at scale will determine how successful it is in the long run.
