On October 15, 2026, Security.com reported that newer Aoki samples analyzed by Symantec Threat Hunter target credentials and session files used by AI coding assistants. The reported targets include Claude, OpenAI Codex, Cursor, Codeium and Google Gemini, according to Security.com’s account.
Which tools and files are named?
The newest sample described also targets tokens for GitHub Copilot, Cursor, Google Gemini, JetBrains AI Assistant and Kiro, along with AWS credentials. Named file examples include .claude.json, Claude desktop cookies, ChatGPT cookies, and OpenAI Codex’s config.toml and .env files.
What could stolen credentials expose?
Security.com describes the targets as credentials, tokens and session files—not source code in general. One qualification: Cursor’s local state database may contain chat context and code snippets.
Stolen credentials or tokens could let someone act under a developer’s identity or use their API quota. They could also expose connected repositories, cloud services or CI/CD pipelines.
