15 JetBrains Plug-ins Stole AI API Keys Across Nearly 70,000 Installs
Updated
Updated · Techzine Europe · Jun 17
15 JetBrains Plug-ins Stole AI API Keys Across Nearly 70,000 Installs
3 articles · Updated · Techzine Europe · Jun 17
Summary
Aikido Security identified 15 JetBrains Marketplace extensions that posed as AI coding assistants while secretly exfiltrating users’ API keys from services including OpenAI, DeepSeek and SiliconFlow.
Nearly 70,000 installs were affected because the plug-ins delivered promised features—chat, code review and testing—while sending saved keys over unsecured HTTP to a hard-coded external server.
Late-October 2025 variants kept reappearing under different names and vendor accounts, with new versions published through June 10, 2026, using near-identical code to evade review.
A paid version deepened the risk: after users paid, the plug-in fetched an API key from the same server, suggesting operators may have reused or resold stolen credentials.
BleepingComputer confirmed the behavior in the latest DeepSeek AI Assist release, which was still listed at publication, underscoring how developer-focused supply-chain attacks can slip past marketplace checks.
With 70,000 developers compromised, can IDE marketplaces ever truly stop malicious plugins that appear to work perfectly?
As AI key theft becomes a lucrative industry, are traditional API keys now too dangerous for developers to use?
Attackers allegedly resold stolen AI access to paying users. Is this the new business model for cybercrime?
2026 JetBrains Plugin Malware: Credential Theft, AI API Key Exposure, and the New Supply Chain Threat
Overview
In June 2026, a major supply chain attack targeted the JetBrains Marketplace through the distribution of credential-stealing plugins. Although such incidents are far less common in JetBrains compared to npm or PyPI, attackers used a sophisticated approach by embedding stealthy data exfiltration logic into plugins that appeared legitimate and fully functional. This allowed the malicious code to evade JetBrains’ manual review process, as revealed by Aikido Security. Since developer machines are extremely high-value targets, this incident highlights the urgent need for stronger security measures and continuous vigilance in the software development ecosystem.