Table of Contents
Modern supply chains run on code, models, and email. The last one is still the easiest way to start a very bad day.
In early September, attackers used a convincing “npm support” email to phish a maintainer, reset 2FA, and push malicious releases to 18 JavaScript packages with roughly 2 billion weekly downloads. The injected code hijacked browser wallet interactions to redirect crypto transactions. The phishing email appeared to come from support@npmjs.help. The pretext looked official, the ask was urgent, and the fallout moved quickly.
Days later, researchers detailed “Model Namespace Reuse,” an AI development supply-chain flaw. If pipelines pull models by name alone from public catalogs, an attacker can reclaim an abandoned namespace, publish a look-alike model, and land code execution when that model is deployed through services such as Vertex AI and Azure AI Foundry. It is a name-trust problem with real RCE consequences.
Why This Should Worry Manufacturers & Logistics Organizations
Manufacturing depends on dense webs of suppliers, integrators, and software partners. A single phishing email to an engineer or vendor manager can lead to poisoned dependencies, tampered build scripts, or fraudulent supplier changes that stall production or reroute payments. That risk multiplies as plants adopt ML models for quality control and forecasting, often pulled from public catalogs.
How The npm Compromise Unfolded
- Initial access: A phishing email posing as npm support prompted a 2FA reset. The maintainer complied. The email address was support@npmjs.help.
- Account takeover: Attackers shipped modified versions of widely used packages.
- Impact: New builds silently intercepted crypto wallet activity in the browser and rewrote payment destinations to attacker addresses.
- Timing: Reports describe the incident unfolding on September 8–9, 2025.
What “Model Namespace Reuse” looks like in practice
- The setup: Teams deploy models by reference, often as Author/ModelName. CI and cloud catalogs keep working even when ownership changes.
- The gap: If the original author deletes or transfers the namespace, an attacker can re-register it, publish a look-alike model, and your pipeline still pulls it by name.
- The outcome: Researchers demonstrated code execution paths when these models are consumed in Microsoft and Google AI services.
The common thread: email trust
Neither incident required a novel exploit to start. Both hinge on human-driven trust. Email is still the first domino adversaries try to tip to reach packages, models, or approvals.
IRONSCALES Could Have Helped
1) Block the phish that stole npm credentials
- Behavioral and relationship analysis at the inbox detects sender anomalies and help-desk look-alikes before users click.
- Intent detection and computer vision flag fake 2FA reset requests and cloned login pages.
- When a developer reports an email, Themis, our agentic AI SOC, clusters similar incidents and pulls every copy in seconds.
2) Disrupt the email pretexts that enable Model Namespace Reuse abuse
Even though namespace reuse is a catalog problem, attackers still reach developers by email with prompts to “migrate,” “verify,” or “upgrade.”
- Vendor impersonation and supply-chain fraud detection block fraudulent supplier and catalog pretexts.
- DMARC management improves signals and stops spoofing at the source.
- Phishing simulations teach engineers to pause on “ownership transfer” and “support ticket” requests, tied to real mailbox threats.
What to Do This Month
- Harden the inbox for developers and vendor-facing teams with inbox-level intent detection and automated remediation.
- Enforce DMARC to stop spoofing at the source.
- Require commit-pinned model references and verified registries, not just names like Author/ModelName.
- Educate engineers on phishing pretexts such as “2FA reset” or “model migration.”
- Use a 90-day scan-back to measure the phishing threats already bypassing your current tools.
Securing the Supply Chain Starts at the Inbox
Every production line, supplier handoff, and logistics chain depends on trusted communication. A single phishing email can derail that trust, introducing risk into vendor payments, production schedules, or quality data.
That’s why manufacturers and supply chain leaders turn to IRONSCALES. Our Adaptive AI and agentic SOC automation are built to stop the credential theft, vendor impersonation, and account takeovers that attackers use to compromise supply chains.
Ask our IRONSCALES experts to show you how we help global manufacturers and their MSP partners secure supply chains every day.
- Request a tailored demo for your manufacturing environment.
- Run a free 90-day scan-back to uncover phishing already in your supply chain communications.
- Speak with our experts about how we protect factories, vendors, and distributors against phishing-driven disruption.
Leave supply chain security up to us. Focus on growing your business.
Explore More Articles
Say goodbye to Phishing, BEC, and QR code attacks. Our Adaptive AI automatically learns and evolves to keep your employees safe from email attacks.