🚨 200,000 GitHub stars in weeks, now a full-scale security disaster.
OpenClaw, once the darling of the open-source AI agent world, is now a hacker's playground:
- 9 public vulnerabilities in five weeks
- 2,200 malicious add-ons infiltrating its marketplace
- 40,000 internet-exposed instances, 93% with authentication bypassed
The real twist? Attackers don't trick users directly, they trick the AI, and the agent tricks you.
Fake setup screens, social engineering from the assistant itself, even macOS malware handed over by the very tool users trust.
This is what happens when viral growth outruns basic security thinking.
Dev teams handed OpenClaw shell access, plugged in emails, Slack, cloud keys... then blindly installed whatever the community offered. Over 40% of audited add-ons had serious flaws. What began as a weekend project became a critical risk overnight.
Now, the malware isn't just targeting humans, it's hijacking trust in the AI agent to do the dirty work.
When your assistant asks for your password, do you ever hesitate?
If OpenClaw is running anywhere in your organisation, assume it's compromised until proven otherwise.
Will we ever put security before shiny new toys?
Would love to hear your thoughts, what's your take?