AI Agents Hacked by First-Ever Info-Stealer Malware

Hackers steal AI agent identities and tokens from exposed OpenClaw instances

  • Information stealers are now targeting AI agent environments, successfully exfiltrating sensitive configuration files from OpenClaw.
  • The stolen files, including authentication tokens and behavioral “souls,” can grant attackers remote access and impersonation capabilities.
  • Hundreds of thousands of OpenClaw instances are reportedly exposed, creating a significant new attack surface for cybercriminals.
  • The project’s virality, with over 200,000 GitHub stars, has attracted heightened security scrutiny and malicious campaigns.

On February 16, 2026, Hudson Rock cybersecurity researchers revealed a first-of-its-kind infection where an information stealer successfully harvested configuration data from an OpenClaw AI agent. This attack, likely perpetrated by a Vidar stealer variant, marks a dangerous evolution in data theft as “the transition from stealing browser credentials to harvesting the ‘souls’ and identities of personal AI agents”.

- Advertisement -

The malware used a broad file-grabbing routine to locate and steal critical files like `openclaw.json`, `device.json`, and `soul.md`. Consequently, attackers could obtain the gateway authentication token, cryptographic keys, and the agent’s core operational principles.

Alon Gal, CTO of Hudson Rock, confirmed the infection details. This incident prompted the OpenClaw maintainers to announce a partnership with VirusTotal to scan for threats, as highlighted by a recent security report.

Meanwhile, the OpenSourceMalware team detailed an ongoing ClawHub malicious skills campaign using a new evasion technique. Security researcher Paul McCarty said this shift shows actors adapting to detection.

Separately, OX Security highlighted security problems with Moltbook, where AI agent accounts cannot be deleted. Furthermore, SecurityScorecard‘s team found hundreds of thousands of exposed OpenClaw instances, creating remote code execution risks.

- Advertisement -

The firm said a single exposed service with high permissions can become a pivot point for attackers. This surge in security concerns follows the project’s massive popularity, which has garnered more than 200,000 stars on GitHub.

✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.

Previous Articles:

- Advertisement -

Latest News

Compound Finance Website Hijacked In Phishing Attack

The front end of lending platform Compound Finance was hijacked on Sunday, redirecting users...

Sen. Lummis pushes for $300 crypto tax exemption

Senator Cynthia Lummis is advocating for a new bill to create a $300 tax...

World of Warcraft scam led Sunny Lu to VeChain

A fraud on Taobao over in-game Gold sparked Sunny Lu's interest in trustless systems.The...

OpenClaw npm Package a Stealthy Data-Stealing Trojan

A malicious npm package posing as an OpenClaw AI installer has been deployed to...

Wall Street Warns of 35% Meltdown Risk Amid Iran War

Wall Street firms like JPMorgan are warning of a potential 10% S&P 500 correction...

Must Read

9 Best Trading Platforms for Crypto Beginners

Many newcomers to the crypto space are looking for platforms to buy, sell and exchange cryptocurrencies. While there are hundreds of crypto exchanges around...