OpenClaw AI agent framework surges in early 2026, raising security worries

In short: OpenClaw, an open-source tool for building AI agents, has surged in early 2026, and security concerns are starting to catch up with the hype.

What's going on

OpenClaw is an open-source framework, meaning its code is public and anyone can copy or improve it. It helps people build AI agents, which are AI programs that can take actions on their own, like a digital assistant that does not just answer questions but also clicks buttons and runs tasks.

The growth numbers are extreme. Reports put OpenClaw at roughly 247,000 to 280,000 GitHub stars, and it briefly gained about 60,000 stars in 72 hours after launch. Its website reportedly hit 27 million monthly visitors, up about 925% from February to March 2026, and it has around 2 million monthly active users.

Developers and companies use it for workflow automation, tool integrations, and even social interaction between agents. One offbeat example was “Moltbook,” a social network where AI agents posted and debated on their own, which Meta acquired on March 10, 2026. In China, OpenClaw also sparked copycat products and a “raise a lobster” craze tied to its logo.

What to watch

Security is the main cloud over the boom. Researchers have warned about issues like prompt injection (tricking an AI with carefully written text) and data exfiltration (getting the system to leak information it should not share). The key question is whether OpenClaw’s community and the companies adopting it can add safeguards fast enough, before real incidents turn excitement into backlash.

Source: Financial Times