Infostealer malware steals OpenClaw AI assistant configuration files
Infostealer malware successfully stole configuration files from an OpenClaw AI assistant victim, including API keys, private device keys, and personality and memory files. The attack represents the first reported infostealer targeting OpenClaw and demonstrates how malware can now extract complex AI agent environments beyond traditional credentials.
What happened
Hudson Rock identified an infostealer malware attack that successfully exfiltrated a victim's entire OpenClaw configuration environment, marking the first reported infostealer attack targeting OpenClaw AI agents.
The malware used a broad file-grabbing routine that swept for specific directory names such as .openclaw, capturing the full operational context of the victim's personal AI agent.
Reported impact
The stolen data included openclaw.json containing the victim's redacted email and a high-entropy Gateway Token, device.json exposing the device's private keys used for secure pairing and signing operations, soul.md files dictating the agent's personality, and memory files storing logs of user activities.
With the extracted private key, an attacker could impersonate the victim's device and access encrypted logs and paired cloud services including Telegram and calendars, granting potential remote access and full control over the AI agent.
Index analysis
This discovery marks a shift from traditional credential theft to the exfiltration of complex AI environments, highlighting growing security risks as AI agents become deeply integrated into professional workflows.
Researchers warned that infostealers may soon add dedicated modules to parse AI agent data, increasing risks for professional workflow systems.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.