aviatrix logoAviatrix
Under Active Attack

Executive Summary

In 2026, Huntress Labs documented a sophisticated social engineering campaign where threat actors weaponized trusted AI platforms including Claude, ChatGPT, and Grok to deliver malware. The FakeAgent campaign exploited Claude Artifacts and shared conversation features to create fake download pages and malicious install guides that appeared legitimate due to hosting on trusted domains like claude.ai, chatgpt.com, and grok.com. Over nine months, attackers delivered SectopRAT, MacSync stealer, and AMOS stealer malware to over 29 organizations by abusing shareable AI content and SEO poisoning techniques. This represents a significant evolution in social engineering tactics as attackers leverage the inherent trust users place in established AI platforms to bypass traditional security awareness training and detection mechanisms.

Why This Matters Now

AI platforms are rapidly becoming integral to daily workflows, creating new attack surfaces that exploit user trust rather than technical vulnerabilities. This trend highlights the urgent need for organizations to reassess security controls around AI tool usage and clipboard-driven execution as traditional domain-based trust models become ineffective.

Attack Path Analysis

MITRE ATT&CK® Techniques

Potential Compliance Exposure

Sector Implications

Sources

Frequently Asked Questions

Attackers leveraged legitimate sharing features like Claude Artifacts and shared conversations to host malicious content on trusted domains, exploiting user trust rather than technical vulnerabilities.

Cloud Native Security Fabric Mitigations and ControlsCNSF

Based on the attack progression modeled above, these are the defensive controls that would constrain each stage.

Aviatrix Zero Trust CNSF would have constrained this AI platform attack by limiting lateral movement through network segmentation and reducing blast radius through workload isolation. The fabric's egress controls and east-west traffic enforcement could have significantly reduced the attacker's ability to move between cloud environments and exfiltrate harvested credentials.

Initial Compromise

Control: Cloud Native Security Fabric (CNSF)

Mitigation: Initial workload compromise would likely still occur, but the fabric's segmentation policies could have limited the scope of accessible network resources and cloud services from the compromised endpoint

Privilege Escalation

Control: Zero Trust Segmentation

Mitigation: RAT deployment may have proceeded, but zero trust policies would likely have constrained the malware's ability to access privileged network segments and cloud service endpoints beyond the initially compromised workload

Lateral Movement

Control: East-West Traffic Security

Mitigation: Lateral movement attempts would likely have been constrained by east-west traffic inspection and policy enforcement, limiting attacker reachability between cloud workloads and sensitive network segments

Command & Control

Control: Multicloud Visibility & Control

Mitigation: C2 communications may have been established, but multicloud visibility would likely have detected and logged anomalous traffic patterns, potentially constraining command execution scope across cloud environments

Exfiltration

Control: Egress Security & Policy Enforcement

Mitigation: Data harvesting could have continued, but egress security policies would likely have constrained outbound data flows, potentially limiting the volume and destinations of exfiltrated credentials and sensitive information

Impact (Mitigations)

Organizational impact would likely have been constrained to isolated network segments, with reduced cross-cloud exposure and limited access to high-value assets protected by segmentation boundaries

Impact at a Glance

Affected Business Functions

  • Information Technology Operations
  • Security Operations
  • End-user Productivity
  • Data Protection
Operational Disruption

Estimated downtime: 2 days

Financial Impact

Estimated loss: $50,000

Data Exposure

Compromised credentials, browser cookies, keychain secrets, SSH keys, cloud access tokens, Telegram sessions, and potential corporate authentication tokens across 29+ organizations. Data includes login credentials for business applications, cloud services, and development environments.

Recommended Actions

  • • Implement egress security and policy enforcement to block unauthorized outbound connections from malicious curl commands and RAT communications to external domains
  • • Deploy multicloud visibility and control systems to detect anomalous interactions with AI platforms and suspicious automation patterns in user behavior
  • • Establish zero trust segmentation with identity-based policies to limit blast radius when credentials are compromised and prevent lateral movement
  • • Configure cloud firewall with URL filtering and AI-powered traffic discovery to identify and block access to malicious AI-hosted content and redirect domains
  • • Enable threat detection and anomaly response capabilities to baseline normal AI platform usage and alert on suspicious download patterns or terminal execution activities

Secure the Paths Between Cloud Workloads

A cloud-native security fabric that enforces Zero Trust across workload communication—reducing attack paths, compliance risk, and operational complexity.

Stop Advanced ThreatsGet a Free Workload Attack Path AssessmentUnder Active Attack?
Cta pattren Image