Security researchers identified at least 15 plugins on the JetBrains Marketplace, across seven publisher accounts, that posed as working AI coding assistants built on DeepSeek and other models while secretly exfiltrating any AI-provider API key a developer entered into the plugin’s settings. The plugins, installed roughly 70,000 times combined, sent stolen keys in plaintext to an attacker-controlled server; researchers suggested the operators may have resold access to the stolen credentials to other users. The campaign had run for about eight months before researchers identified the shared malicious code and notified JetBrains.