Security researchers at Huntress have documented what they describe as one of the first observed cases of "vibe-coded" malware in the wild, in which an intruder used an AI-generated PowerShell script to enumerate a victim's Active Directory environment. The finding signals a subtle but meaningful shift in how lower-skilled threat actors assemble their tooling.
June 3, 2026 · Threat Research
"Vibe-Coded" Malware Hits the Wild — an AI-Written Script Maps Active Directory
A low-skill intruder prompted an AI until it produced a working PowerShell reconnaissance tool, then ran it against a live network. The code was disposable — but its behavior gave it away.
5-stage
fallback cascade to locate domain controllers
~30 min
from RDP login to running the AI-built script
1
one-off script — no reusable signature to hash
Signature vs. Behavior detection
AI can rewrite code syntax endlessly to dodge static hashes — but the underlying enumeration produces the same system calls every time. The taller the column, the more attempts it survives.
Fails fast
Static hash / signature detection — beaten by disposable code
Still catches it
Behavioral telemetry — same footprint no matter how code is written
Inside the attack — what the script did
1 · Stage
Tools dropped in C:\ProgramData\ over stolen-credential RDP
→
2 · Enumerate
Users, computers, groups, OUs, subnets & trusts → CSV
→
3 · Package
Timestamped report dir + HTML summary, zipped up
→
4 · Exfil
s5cmd.exe (S3) + SharpShares.exe for share enumeration
The tell — machine authorship left in plain sight
Unedited AI artifacts betrayed the script's origin — hallmarks of iterative natural-language prompting, not hand-crafted code.
Boastful auto-generated title
"100% Working... FULLY FIXED"
Placeholder left in the code
Server1.HR.local
⚠ Why it matters
Vibe coding lowers the barrier: even mediocre operators can spin up bespoke, single-use scripts, raising the volume of tailored, hard-to-fingerprint intrusions.
✓ Why it's not the apocalypse
The playbook is unchanged — only speed and personalization shift. Behavior-based detection still flags the activity, so defend around what attackers do , not what their code looks like.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…