A threat actor claiming to be a bug bounty hunter likely used a large language model to develop malware that steals credentials from software developers, according to a CrowdStrike analysis published this week. The cybersecurity firm assessed with high confidence that the developer behind PhantomRaven, a JavaScript-based information stealer, wrote the malicious code using AI based on verbose comments, placeholder code, and statistical token-analysis patterns. The threat actor has been active since November 2022 and claims to have collected bounties from at least nine organizations across technology, retail, and hospitality.
PhantomRaven was initially identified by Koi Security and DCODX in late October 2025 during a slopsquatting and typosquatted campaign where more than 100 malicious packages were uploaded to npm to harvest authentication tokens, CI/CD secrets, and GitHub credentials from developers' machines. The software supply chain attack used these packages as a cover to pull a remote dynamic dependency from an external server so the libraries themselves wouldn't be flagged by security tools. Once installed, the malware scans developer environments for email addresses, gathers CI/CD environment information, collects system fingerprints including public IP addresses, and sends results to an attacker-controlled server. It's also designed to collect runtime details, current date and time, username and email addresses from Git and npm configurations, plus CI/CD environment variables for GitHub Actions, GitLab CI, Jenkins, and CircleCI. At least two different npm user accounts maintained by the operator pushed packages containing PhantomRaven, though both accounts are no longer accessible. Other online identities linked to the same operation include jpd12, jpd13, npmhell, npmpackagejpd, npmtestdharsh, jpdhackerone11, and packagedharsh.
CrowdStrike has not observed information stolen from the malware appearing on stealer log shops, indicating "the operator likely uses the information stealer solely to identify bug bounty opportunities," according to the report. Security researcher Maddie Stewart noted that in August 2025, the threat actor claimed to have discovered a remote code execution vulnerability via a malicious npm package they published, explaining they had compromised the target machine and executed their preinstall script to achieve RCE. Evidence has also emerged that the threat actor attempted to push packages to the Python Package Index repository containing code for an information stealer that exhibits similarities with PhantomRaven.
The likely use of LLM technology to generate the malware highlights how threat actors are increasingly adopting AI in their operations, compressing the time and effort required to execute such campaigns. "Most criminal actors [...] rent commodity tools or operate their own proprietary malware; however, this threat actor has likely developed their proprietary PhantomRaven to compromise company assets and then used these compromises as leverage to claim rewards from reputable disclosure programs," CrowdStrike stated. The approach represents an unusual method where stolen credentials aren't sold on underground markets but instead weaponized to exploit legitimate security reward programs for financial gain. The case demonstrates how AI tools can lower technical barriers for cybercriminals while simultaneously creating new attack vectors that abuse trusted software ecosystems and corporate vulnerability disclosure processes.