ANY.RUN Ties 1,258 Sandbox Runs to Fake AI Brand Impersonation Malware Hitting US Targets

The campaigns rotate domains and AI brands while reusing the same living-off-the-land sequence: clipboard copy of a PowerShell command, retrieval of an HTA payload, and staged execution that evades static filters. ANY.RUN sandbox traces show consistent final-stage credential theft and session hijacking across variants. US companies appear as repeated targets because they rely on the same productivity tools the lures mimic.

Evidence from the sandbox runs reveals operational friction for MSSPs: initial indicators change faster than signature updates, forcing repeated deep dives into similar execution graphs. This pattern matches prior waves of fake software installers observed in 2024-2025 contractor procurement leaks, where domain flux and brand abuse outpaced shared threat feeds. No state attribution is supported by the technical artifacts alone.

The pressure concentrates on Tier-1 triage and cross-customer correlation. When the same behavioral sequence appears under different customer policies, escalation rates rise and MTTR stretches. MSSPs that fail to build reusable detection logic on the HTA-to-PowerShell pivot will see SLA breaches as volume grows.

Next steps require shared behavioral rulesets focused on clipboard-initiated mshta activity and rapid domain pivoting rather than brand-specific IOCs. Vendors publishing new AI tooling should expect their names to be abused within weeks of public release.