Hackers use artificial intelligence to steal secrets - trust tools to attack developers

Hackers are now leveraging large language models (LLMs) for advanced cyberattacks, leveraging trusted AI tools to extract sensitive data. In a recent intrusion, the attacker hacked into the npm account of the widely used Nx package, implanting malware that utilizes artificial intelligence assistants such as the Gemini CLI or Claude Code CLI to search and steal cryptocurrency wallet passwords and API keys. The stolen data is uploaded to GitHub in base64 encoding, circumventing traditional security measures. This incident highlights the increasing risks of AI-driven attacks, as attacks are more deceptive and more difficult to detect, and it is urgent to strengthen security in the open source ecosystem.