AI-Powered Ransomware Executes First Real-World Attack

Cybersecurity researchers have uncovered what is believed to be the first real-world ransomware attack executed by an autonomous large language model (LLM). The AI agent, powered by a threat actor dubbed JadePuffer, compromised an internet-exposed instance of Langflow connected to Alibaba Nacos and MySQL databases in late June 2026.

The LLM ran the full playbook, including reconnaissance, credential harvesting, lateral movement, privilege escalation, and database encryption on a MySQL/Nacos server. The attack destroyed 1,342 configuration items in the process.

The AI was specifically programmed to scan for cryptocurrency wallets and seed phrases, and it also hunted for API keys from major cloud providers. However, the ransom note left behind demanded Bitcoin payment to an address pulled straight from Bitcoin documentation, which is a known example address.

Interestingly, the encryption key generated by the AI was non-functional and was never transmitted back to the attacker, making decryption impossible even if the victim had wanted to pay.