The Hades malware campaign is targeting scientific and machine-learning development packages by using adversarial prompt-injection attacks to bypass automated security scanners and steal sensitive developer credentials and keys.

Key Points

• The Hades campaign currently impacts 37 Python and 106 JavaScript packages, including several typo-squatting variants of popular libraries.
• Attackers are embedding prompt-injection instructions in JavaScript comments to trigger AI safety filters, causing scanners to abort before analyzing malicious payloads.
• The malware now exfiltrates a wide range of sensitive data, including AWS credentials, SSH keys, Kubernetes tokens, and .env files.
• New evasion techniques include splitting payloads across multiple packages and delaying execution until the package is initialized rather than installed.
• The malware includes self-destruct mechanisms that trigger if it detects a sandboxed environment or security analysis tools.

Why it Matters

This campaign highlights a critical vulnerability in the growing reliance on AI-driven security tools for scanning software supply chains. By exploiting the safety protocols of these models, attackers can effectively blind automated defenses, leaving development environments exposed to credential theft and unauthorized access.