Fake 'Bug Bounty Hunter' Used AI-Written Malware to Raid npm Developer Secrets

Security researchers at CrowdStrike have linked a financially motivated threat actor to PhantomRaven, a JavaScript-based information stealer distributed through more than 100 malicious npm packages. Analysis of the code, including verbose comments and placeholder text, suggests it was likely written with help from a large language model.

First identified by Koi Security and DCODX in late October 2025, the campaign used typosquatted and AI-hallucinated package names to trick developers into installing infected libraries. Rather than embedding malicious code directly in the packages, the attacker fetched a hidden remote dependency after installation, helping the packages evade security scanning. Once active, the malware collected email addresses, CI/CD environment details, system fingerprints, public IP addresses, and credentials from Git and npm configurations across platforms like GitHub Actions, GitLab CI, Jenkins, and CircleCI.

CrowdStrike found the operator has been active since November 2022 and claims to be a legitimate bug bounty hunter, having reportedly collected bounties from at least nine companies in technology, retail, and hospitality. Notably, stolen data has not appeared in criminal stealer log marketplaces, suggesting it may be used to identify vulnerabilities for bounty claims rather than sold outright. The npm accounts behind the campaign have since been taken down.