Blog
Product updates, security research, and field lessons from the OpenGuardrails team.
February 21, 2026
OpenGuardrails has identified a new, rapidly spreading malware campaign targeting the OpenClaw ecosystem through the ClawHub skill community. We are naming this threat Clawhub.Trojan.LiuComment.
February 20, 2026
When Claude Code Security was announced on February 20, 2026, global cybersecurity stocks dropped almost immediately. The market understood: the first principle has not changed, but the cost law governing the entire industry just did.
February 4, 2026
AI agents now read emails, browse web pages, and process documents autonomously. But what happens when attackers hide malicious instructions inside those long-form contents? Today, we release OG-OpenClawGuard v1 — an open-source Guard Agent plugin for OpenClaw that detects and blocks indirect prompt injection attacks in real time.
January 29, 2026
Personal AI assistants now execute shell commands, browse the web, and send messages — but most users have zero visibility into what their agents can actually do. Today, we're introducing OG Personal, the first guard agent designed specifically for personal AI assistants.
January 23, 2026
Large Language Models have become the second brain of modern enterprises. But in real enterprise environments, one uncomfortable question keeps surfacing: do we actually know how much sensitive data is being sent to external LLMs — unintentionally?
December 27, 2025
A global call to shape AI Runtime Security together. OpenGuardrails announces the AI Runtime Security Management System (AI-RSMS) — an open, community-driven standard draft focused on securing AI systems during runtime.
December 16, 2025
OpenGuardrails 4.5.0 introduces Direct Model Access—a privacy-first feature that lets enterprises quickly deploy private POCs by pointing to our SaaS models without logging any data. Deploy locally, access models remotely, keep everything private.
November 5, 2025
Help Net Security spotlighted how OpenGuardrails unifies prompt-attack defense, moderation, and sensitive-data protection in one configurable, scalable, open-source guardrail stack ready for production workloads.