FakeAgent Campaign Abuses Trusted Claude.ai Domain to Deliver SectopRAT Infostealer
Attackers in the FakeAgent campaign exploited user trust in Anthropic's legitimate claude.ai domain by hosting malicious artifacts on the platform itself, effectively weaponizing a reputable service to bypass security controls and increase victim credibility. This is a classic supply chain trust abuse: because the payload was served from a known, allowlisted domain, traditional reputation-based defenses failed to flag it. At least 29 organizations were compromised within a 48-hour window, demonstrating how rapidly malvertising campaigns can scale when leveraging trusted infrastructure. The SectopRAT infostealer deployed in this campaign can harvest credentials, browser data, and sensitive system information, making the downstream impact significant. This case underscores that domain reputation alone is not a sufficient security signal — content and behavior must also be scrutinized.
Tactical Insight
Immediate actions
- Block or sandbox downloads originating from public AI artifact-hosting platforms (e.g., claude.ai, similar services) until content can be verified by security teams.
- Deploy endpoint detection and response (EDR) rules specifically targeting SectopRAT indicators of compromise (IOCs) published by Huntress across all managed endpoints.
- Alert and re-educate users who searched for or downloaded Claude Desktop during July 21–22, 2026 to submit their devices for forensic review.
Long-term improvements
- Implement application allowlisting so only pre-approved, signed software installers can execute, regardless of the hosting domain's reputation.
- Establish a formal vetting process for any AI-generated or AI-hosted artifacts before they are permitted in the enterprise environment.
- Integrate threat intelligence feeds that monitor malvertising campaigns and trusted-domain abuse patterns into your SIEM for proactive alerting.
Detection measures
- Enable DNS and web proxy logging to flag anomalous download activity from AI platform domains and correlate it with executable file types.
- Deploy behavioral analytics to detect SectopRAT post-infection patterns such as credential harvesting, C2 beaconing, and browser data exfiltration.
- Conduct regular threat hunts using the FakeAgent campaign IOCs to identify any previously undetected compromises across the environment.