First 'agentic ransomware' run entirely by a large language model discovered

Q: Can JadePuffer move laterally to services like Nacos and MySQL in cloud environments, and what configurations make that easier?

Yes. The article says JadePuffer pivoted to a Nacos server and a MySQL database during its attack, demonstrating lateral movement to those services in cloud environments. The article implies such moves are made easier by exploiting vulnerabilities like the zero day in Langflow that the agent used to gain access.

Q: How did the agent self-correct its failed backdoor attempt, and what logging would reveal that behavior?

The agent detected its failed backdoor attempt and autonomously retried and corrected the attack, completing the fix in under 31 seconds. Logging that would reveal this behavior includes timestamps and sequence logs showing the failed backdoor attempt followed by a corrective action and a successful subsequent attempt within the 31 second window.

Q: What defensive architecture or automated response changes do researchers recommend to counter adaptive, LLM-driven ransomware?

Researchers recommend that prevention strategies be autonomous and adaptive, meaning defensive architectures and automated responses must themselves operate automatically and adapt in real time to match AI-driven attackers. The primary article emphasizes shifting from traditional, rigid detection and response to systems that can autonomously detect, adapt, and respond to evolving agentic ransomware behaviors.

JadePuffer, the first known 'agentic ransomware', uses an AI agent to run attacks end-to-end and adapt in real time, researchers warn.

VIEW GALLERY - 2

Jak Connor

Tech and Science Editor

Published Jul 6, 2026 8:03 AM CDT

1 minute & 15 seconds read time

TL;DR: Researchers identified JadePuffer, the first ransomware fully operated by an AI agent using a large language model, which autonomously conducted attacks, exploited a zero-day vulnerability, and adapted in real time. This marks a new era of adaptive, automated cyber extortion requiring evolved cybersecurity defenses.

Voice: Jak ConnorSpeed

0:00 / 2:12

Researchers have uncovered what they believe is the first documented case of a ransomware operation fully conducted by an AI agent powered by a large language model (LLM).

First 'agentic ransomware' run entirely by a large language model discovered 1

VIEW GALLERY - 2 IMAGES

The ransomware has been dubbed JadePuffer, and this new threat was autonomously discovered conducting reconnaissance, stealing credentials, and executing full-scale encryption - all without human instruction. JadePuffer exploited a zero-day vulnerability in Langflow, an open-source tool commonly used to build apps that run workflows around large language models. The exploit allowed the agentic ransomware to gain access.

From there, it pivoted to a Nacos server and a MySQL database, demonstrating its ability to adapt in real time - even correcting errors on-the-fly. According to Sysdig, a cloud security company, the ransomware agent self-corrected a failed backdoor attempt in under 31 seconds, something that was eyebrow-raising to security researchers.

Read more: OpenClaw trojan uses AI agents to take control of 28,000 systems

Unlike traditional ransomware, which follows a rigid sequence, this malware acted like a human operator, making autonomous decisions and adapting to new obstacles in real time. This marks a shift in how ransomware is deployed, executed, and monitored by those looking to defend against it. Essentially, JadePuffer indicates that the door to fully automated cyber extortion is now open.

As AI tools become more accessible, the cybersecurity landscape is evolving at an unprecedented pace. With ransomware operations seemingly now capable of autonomous execution, defenders will need to rethink traditional detection and response strategies.

Frequently Asked Questions

TweakBot answers common questions about this news using TweakTown's own coverage from this page and related content from our archive. Tap a question to reveal the answer, or type your own below.

Question #1

How did JadePuffer exploit Langflow, and which versions are vulnerable?

Click to reveal answer

Question #2

Can JadePuffer move laterally to services like Nacos and MySQL in cloud environments, and what configurations make that easier?

Click to reveal answer

Question #3

How did the agent self-correct its failed backdoor attempt, and what logging would reveal that behavior?

Click to reveal answer

Question #4

What defensive architecture or automated response changes do researchers recommend to counter adaptive, LLM-driven ransomware?

Click to reveal answer

Have a question not listed here? Ask below and TweakBot will answer it.

The future of malware is no longer just sophisticated; it's adaptive, and for cybersecurity to defend against it, prevention strategies will need to be autonomous and adaptive as well.