JetBrains disclosed a supply chain attack involving 15 malicious third-party plugins on JetBrains Marketplace that stole AI provider API keys (OpenAI, DeepSeek, SiliconFlow, etc.). The plugins posed as legitimate AI utilities, silently installed a JVM-wide X509TrustManager to bypass TLS warnings, then exfiltrated API keys as plaintext JSON to a hardcoded C2 IP address. JetBrains removed all 15 plugins, permanently banned 7 publisher accounts, and triggered a remote kill-switch to disable the plugins in users' IDEs. Affected developers should immediately revoke and reissue any API keys entered into these plugins, check provider dashboards for anomalous usage, and block the C2 IP. JetBrains is hardening its ingestion pipeline to flag raw HTTP endpoints, hardcoded IPs, and unauthorized TLS weakening. The post also recommends adopting the Agent Client Protocol (ACP) as a safer alternative to unsandboxed marketplace plugins.
Table of contents
Technical analysis: How the malicious activity occurredHardening our vetting pipelinesRequired actions and step-by-step remediationMoving Forward Safely: The ACP Protocol12.9K Impressions1 Comment