In 2026, the open-source ecosystem faced a new wave of attacks on npm packages, notably through the Mini Shai-Hulud campaign and the Miasma incident. Attackers compromised packages within the @redhat-cloud-services namespace, injecting code designed to steal cloud tokens, credentials, and CI/CD infrastructure secrets. This case highlights the critical importance of supply chain security, especially in an era of rapid AI integration.
Market implications
Modern AI solutions are deeply integrated with corporate data, APIs, and cloud platforms. According to the MITRE ATLAS framework, AI security encompasses not only the model itself but also its entire supporting infrastructure. Compromising even a single npm package can grant attackers access to machine learning environments, risking data leaks, training data poisoning, or model compromise.
In the latest edition of the OWASP Top 10 for LLM Applications, third-party component risks are categorized separately as Supply Chain Vulnerabilities (LLM05). To mitigate these threats, the National Institute of Standards and Technology (NIST), under its AI Risk Management Framework (AI RMF 1.0), recommends implementing four core functions: Govern, Map, Measure, and Manage.
Recommendations
To defend against supply chain attacks, organizations must implement comprehensive security measures:
- Dependency control: regular use of SCA solutions to analyze third-party libraries.
- CI/CD protection: automated code scanning (SAST/DAST) and digital signature verification.
- Least privilege principle and MFA: restricting access for services and enforcing multi-factor authentication.
- AI integration audits: testing models for resilience against specific attacks (prompt injection, adversarial attacks).
According to Gartner analysts, demand for specialized AI security platforms is set to grow rapidly. However, robust protection is only possible by combining technical tools, clear governance processes, and regular staff training.
Prepared by a Software Ukraine member. Original publication.