The Mini Shai‑Hulud Supply‑Chain Attack Explained

The TanStack Compromise

One of the largest targets was the TanStack ecosystem, which includes popular JavaScript libraries used in many web applications.

Within minutes, attackers published 84 malicious versions across 42 @tanstack/* npm packages through the project's legitimate release pipeline.

Installing these versions triggered malicious code during npm lifecycle hooks that downloaded and executed a credential‑stealing payload.

Security researchers later reported that the campaign expanded rapidly:

• 170+ packages across npm and PyPI were affected
• 403 malicious package versions were identified during the largest wave
• Ecosystems linked to UiPath, Mistral AI, OpenSearch, and Guardrails AI were also impacted

This made Mini Shai‑Hulud one of the most widespread open‑source supply‑chain compromises of 2026.

How the Malware Worked

The malicious packages used standard package‑manager hooks (such as preinstall) to execute code when developers installed them.

Once executed, the malware attempted to collect developer credentials from the infected system. Researchers observed it targeting:

• GitHub personal access tokens
• npm publishing tokens
• Cloud provider credentials (AWS, Azure, GCP)
• Kubernetes secrets
• HashiCorp Vault tokens
• CI/CD environment variables
• SSH keys and developer configuration files

These credentials could then be used to compromise additional repositories and publish more infected packages, allowing the worm to propagate through developer infrastructure automatically.

How Two OpenAI Employee Devices Were Affected

During the broader campaign, malicious TanStack npm packages were installed on two OpenAI employee devices within the company’s corporate environment.

OpenAI said the attackers conducted unauthorized access and credential‑focused exfiltration activity on those systems. However, the company reported that its investigation found:

• No evidence that customer data was accessed
• Only limited credential material was exposed

Public reporting does not provide the exact infection chain or identify which specific TanStack package versions triggered the compromise.

Why OpenAI Rotated macOS App Certificates

Following the incident, OpenAI updated its guidance for organizations that allowlist OpenAI macOS applications.

The company confirmed that the Apple Developer signing identity used by its macOS apps changed as part of its response, requiring some enterprise security policies to be updated.

Important details for administrators:

• The Apple Developer Team ID remains the same: 2DC432GLL2
• Allowlisting by Team ID may still work without changes
• Policies that verify certificate fingerprints or signing organization names may need updating

Organizations using macOS application allowlisting should verify they trust the current OpenAI signing identity rather than older certificates.

Wider Impact on npm and PyPI

The attack highlighted a broader risk in modern software development: automated package publishing pipelines can become attack surfaces.

Unlike traditional supply‑chain compromises that steal maintainer credentials, Mini Shai‑Hulud exploited:

• CI/CD automation
• Trusted publishing identities
• Package registry trust models

Because of this design, malicious releases could appear legitimate and spread quickly across ecosystems.

By the end of the largest wave:

• More than 170 npm and PyPI packages had been compromised
• Hundreds of malicious versions were published
• Multiple developer ecosystems were temporarily affected

What Developers and OpenAI Users Should Do

1. Audit dependencies installed around May 11–12, 2026

Developers should review builds and dependency installs during the window when malicious versions of TanStack packages were published.

2. Remove and reinstall affected packages

If a compromised version was installed:

• Remove the dependency
• Install a known‑good version
• Rebuild the project from clean sources

3. Rotate developer credentials

Because the malware targeted credential material, rotate any secrets that might have been exposed, including:

• GitHub tokens
• npm or PyPI publish tokens
• cloud provider credentials
• CI/CD secrets

4. Review GitHub Actions publishing workflows

Projects should audit CI/CD pipelines that publish packages and verify that OIDC trusted publishing configurations are tightly restricted.

5. Update macOS allowlists for OpenAI apps

Organizations running OpenAI desktop tools should ensure their macOS policies trust the updated signing identity used by OpenAI apps.

What This Incident Shows About Modern Supply‑Chain Security

Mini Shai‑Hulud illustrates how modern attacks increasingly target developer infrastructure rather than production systems directly. By compromising the tools and pipelines developers rely on, attackers can distribute malware through trusted software dependencies.

The campaign also demonstrated a troubling new reality: even packages with valid provenance attestations and legitimate publishing signatures can still be malicious if the underlying build pipeline is compromised.

For teams that rely heavily on open‑source dependencies, monitoring dependency changes, isolating build environments, and rotating credentials quickly after incidents have become essential parts of software security.