Ant Group Open-Sources Two AI Security Frameworks
量子位 · wechat · 2026-07-12
The article discusses new security risks introduced by agents and multimodal models: risks are no longer just content-based but extend to tool usage, code execution, perception, and behavior. Using cases like Claude Code and OpenClaw, the author argues that traditional "patch-based" security is insufficient, calling for more foundational security frameworks.
It highlights two open-sourced frameworks from Ant:
- SingGuard-NSFA: Designed for agent security, supporting both generative reasoning and discriminative classification. It sets checkpoints at both request interception and response fallback, emphasizing explainable and scalable risk management.
- SingGuard: Designed for multimodal LLMs, treating safety rules as runtime inputs, supporting fast/slow thinking switches, and boosting multi-rule parallel review efficiency via RI-Mask.
The frameworks have achieved strong results in multiple evaluations. Ant's ongoing efforts in vulnerability discovery, ClawAegis, and third-party certifications point to a broader vision of "building security infrastructure for the AI era."
More from Safety
- AI Security Institute says every tested model tried to cheat in cyber evaluations — connoraxiotes · 2026-07-21
- Congressional brief warns AI could speed biology research while creating new biosecurity risks — sebkrier · 2026-07-21
- AI Companies Are Buying Tons of Old Books Because They're Free of AI Slop — 404 Media · 2026-07-21
- A simple standup question exposes who owns AI model approval in customer workflows — YvesMulkers · 2026-07-21
- Anthropic says frontier models showed harmful behavior in tool-rich simulations — gerardsans · 2026-07-21
- Cisco releases Antares small models to localize code vulnerabilities — aminkarbasi · 2026-07-21