What Is OpenClaw and Why Is It Trending?

Originally published: February 12, 2026 | Updated: April 9, 2026

The open-source AI agent with 351,000+ GitHub stars — and why Hong Kong’s government restricted it

OpenClaw「龍蝦」is an open-source autonomous AI agent created by Austrian developer Peter Steinberger. Originally named Clawdbot, then briefly Moltbot, the project was renamed following trademark concerns from Anthropic. Steinberger has since joined OpenAI, while the OpenClaw codebase transferred to an independent foundation.

Unlike traditional chatbots that answer questions inside a browser, OpenClaw runs locally on a user’s device and autonomously executes tasks. It connects to messaging platforms including WhatsApp, Telegram, Slack, Discord, iMessage, WeChat, LINE, and Microsoft Teams. It can manage email, browse the web, control files, schedule appointments, and execute multi-step workflows — all through natural language commands.

By April 2026, OpenClaw had surpassed 351,000 GitHub stars and 70,000 forks — more than React accumulated in over a decade — making it the most-starred software project in GitHub history. It has 3.2 million monthly active users and 38 million monthly website visitors.

In the Chinese-speaking world, OpenClaw is widely known as「龍蝦」(lobster) after its red lobster mascot, and the practice of setting it up is called「養龍蝦」(raising the lobster). China’s traffic to OpenClaw grew 1,436% month-over-month in early 2026, driven by official Tencent WeChat and QQ Bot integrations.

Key OpenClaw facts (updated April 2026):

Is OpenClaw Safe for Enterprise Use?

Most cybersecurity experts advise against deploying OpenClaw in enterprise environments without significant security hardening.

Palo Alto Networks identified OpenClaw as a “lethal trifecta” of risk: access to private data, exposure to untrusted content, and the ability to perform external actions while retaining persistent memory (CNBC, February 2, 2026).

Cisco described it as “groundbreaking” in concept but an “absolute nightmare” for security (Information Age, February 2026).

The Register called the project a “security dumpster fire” after documenting a one-click remote code execution vulnerability (CVE-2026-25253, CVSS 8.8) and two command injection vulnerabilities (The Register, February 3, 2026).

Since February 2026, the security picture has escalated:

• 9 CVEs disclosed in 4 days
• 135,000+ OpenClaw instances found exposed on the public internet across 82 countries, with 15,000+ directly vulnerable to remote code execution
• Hundreds of malicious extensions identified in ClawHub, with CrowdStrike warning that OpenClaw can be “commandeered as a powerful AI backdoor agent”
• 36% of all ClawHub skills found to contain prompt injection vulnerabilities

Steinberger has acknowledged these risks: “It’s a free, open source hobby project that requires careful configuration to be secure. It’s not meant for non-technical users.”

Hong Kong’s Response: Government Restriction and HKCERT Guidance

In March 2026, Hong Kong’s authorities took concrete steps in response to OpenClaw’s risks.

The Digital Policy Office（數字政策辦公室）and the Government Computer Emergency Response Team issued guidance to civil servants, directing that OpenClaw should not be installed on computers connected to the government’s internal network at this stage. 

Innovation and Technology Secretary, Sun Tong stated that while OpenClaw provides AI-assisted applications, it may also bring risks including excessive permissions, potential data leakage, and system intrusion.

HKCERT（香港電腦保安事故協調中心）issued a public advisory assessing that OpenClaw’s risk profile significantly exceeds that of typical chatbot AI, and provided five security recommendations.

These developments signal that while AI agents represent a genuine business opportunity, the approach matters as much as the technology. Enterprise environments — particularly those handling customer data, operating in regulated industries, or communicating across platforms like WeChat and WhatsApp — require a fundamentally different approach than installing an open-source hobby project on a company computer.

The Bigger Question: Is Your Data Ready for AI Agents?

The security risks are fixable engineering problems. The deeper challenge is structural — and it goes beyond customer data.

AI agents do not create intelligence from nothing. They amplify whatever data, knowledge, and processes they access. Cisco’s AI Readiness Index found that only 13% of organisations globally have the infrastructure to make AI agents effective. Gartner predicts more than 40% of agentic AI projects will be cancelled by 2027 due to escalating costs, unclear value, or insufficient governance.

The OpenClaw hype focuses on personal productivity. But for enterprises, AI agent readiness spans three distinct challenges:

Customer-facing: Customer data is scattered across CRM, email, WhatsApp, WeChat, and support tools. No single view of the customer exists. AI agents deployed here personalise with bad data and automate guesswork.

Internal operations: Business workflows are documented but still run manually — data entry, invoice processing, approval routing. AI agents need structured, integrated processes to automate effectively.

Knowledge and data: Institutional knowledge lives in PDFs, email threads, and people’s heads. Data warehouses hold answers but require technical skills to access. AI agents cannot use knowledge they cannot find.

Turning scattered data into business decisions is where competitive advantage will be won.

Is your company ready for AI Agent deployment? Talk to us.

What Should Your Enterprise Do Next?

OpenClaw has demonstrated real demand for AI agents that act autonomously across messaging, email, and business systems. But HKCERT has assessed that its risk profile exceeds typical chatbot AI, and the HK OpenClaw has demonstrated real demand for AI agents that act autonomously across messaging, email, and business systems. But HKCERT has assessed that its risk profile exceeds typical chatbot AI, and the HK government has directed that it should not be installed on government-connected computers at this stage. For enterprises, the question is not whether AI agents are useful — it’s how to deploy them responsibly.

Master Concept helps enterprises across Hong Kong, Singapore, and Taiwan deploy AI across two pillars:

For customer-facing AI — DAL (Data & AI Lab)

If your challenge is understanding customers better, personalising engagement, and unifying customer data across WhatsApp, WeChat, CRM, and support tools:

Business leaders: Read The 87% Problem: Why Most Hong Kong Enterprises Will Fail at AI — find out if you’re in the 13% that are AI-ready, or the 87% that need to act now.

Digital and operations heads: Read 從「養龍蝦」到企業級 AI 智能體 — the enterprise-grade approach to AI-powered customer engagement.

Learn the framework: Read What Is Customer Intelligence? — the discipline of unifying customer data, understanding it through analytics, and acting on it in real time.

For internal operations AI — MasterAI

If your challenge is slow data access, buried knowledge, or manual processes that consume your team’s time:

Need instant insights from your data?

InsightAI lets anyone ask questions in natural language and get instant visualizations from your data warehouse — no SQL, no waiting for the data team.

Need a searchable internal knowledge base?

KnowledgeAI builds an AI-powered knowledge base from your documents — with citations and smart retrieval. Your team finds answers in seconds, not hours.

Need workflow automation?

ProcessAI automates business processes with AI — from invoice processing to procurement — with human review at critical decision points.

Explore the full platform:

Read What Is MasterAI? — three products, one platform, powered by AWS. Go live in weeks, not months.

Getting started

IT managers exploring AI agents after seeing the 養龍蝦 trend: Join the 馴服龍蝦 workshop — Master Concept’s hands-on course for secure AI agent deployment.

Singapore enterprises: Read Singapore’s AI Agent Paradox — guidance specific to MAS-regulated environments.

Not sure where to start? Take the AI Agent Readiness Scorecard — 10 questions, 3 minutes, instant results across customer intelligence, operational AI, and data foundations.

Or talk direct to our AI deployment team

Frequently Asked Questions

What is OpenClaw?

OpenClaw（龍蝦）is an open-source autonomous AI agent created by Peter Steinberger, who has since joined OpenAI. It runs locally on a user’s device, connects to messaging platforms including WhatsApp, Telegram, WeChat, and LINE, and autonomously executes tasks. As of April 2026, it has 351,000+ GitHub stars, 3.2 million monthly active users, and 44,000+ community-built skills.

Is OpenClaw safe for enterprise use?

Most cybersecurity experts advise against enterprise deployment without significant security hardening. Nine CVEs were disclosed in four days, 135,000+ instances were found exposed on the public internet, and hundreds of malicious extensions were found in its skills marketplace. In Hong Kong, HKCERT issued security guidance, and the government directed that it should not be installed on government-connected computers at this stage.

Did Hong Kong ban OpenClaw?

No. The Hong Kong government directed that OpenClaw should not be installed on computers connected to the government’s internal network at this stage. This is an internal IT security measure for government systems, not a public ban. HKCERT has separately issued public security guidance noting that OpenClaw’s risk profile exceeds that of typical chatbot AI.

Why do enterprise AI agent projects fail?

The primary cause is insufficient data infrastructure, not AI model limitations. This spans three areas: fragmented customer data (no single customer view), inaccessible knowledge (buried in PDFs, emails, and people’s heads), and manual processes (documented but not automated). Gartner predicts over 40% of agentic AI projects will fail by 2027 due to escalating costs, unclear value, or poor data governance. Cisco found only 13% of organisations globally are fully AI-ready.

What is Customer Intelligence?

Customer Intelligence is the business capability of unifying customer data from every touchpoint, understanding it through analytics, and acting on it in real time. It is built in three layers: Signal (data capture and unification), Insight (analysis and pattern recognition), and Action (personalised engagement and automation). Learn more about how this applies to Hong Kong enterprises →

What is MasterAI?

MasterAI is Master Concept’s enterprise AI platform powered by AWS. It includes three products: InsightAI (natural language data analytics — ask questions, get instant visualizations), KnowledgeAI (AI-powered knowledge base with document retrieval and citations), and ProcessAI (workflow automation with human review at decision points). SOC 2 certified, enterprise SSO, data residency controls. Go live in 1-4 weeks. Learn more about MasterAI →

How do I know if my company is ready for AI agents?

Assess three zones:
(1) Customer Intelligence — do you have a unified customer view and real-time signals?
(2) Operational AI — are your workflows documented and integrated enough for automation?
(3) Knowledge and data — can your team access insights and institutional knowledge without technical skills or waiting?

Take the 3-minute AI Readiness Scorecard →

DAL (Data & AI Lab) helps enterprises across Hong Kong, Singapore, and Taiwan close the gap between data rich and intelligence ready. Book a Customer Intelligence Assessment →