TechCrunch · AIJul 7, 2026, 07:56 UTC+8

The 'first' AI-run ransomware attack still needed a human to pick the target

The ‘first’ AI-run ransomware attack still needed a human

Sysdig last week reported the first 'agentic ransomware' called JadePuffer, where an AI agent handled the full technical execution. Sysdig's Michael Clark clarified Monday that a human still chose the victim, set up the infrastructure, and supplied stolen credentials. The AI ran the attack steps but wasn't fully autonomous.

Why it matters: Sysdig corrected its own narrative from last week, shifting from 'AI autonomously ran ransomware' to 'human set the stage, AI executed.' That detail reversal carries more signal than the original report. Score capped at 72 because it's still a vendor case study, not a product ...

Read the original ↗Export Markdown

Take

Last week Sysdig framed JadePuffer as the first ransomware attack fully executed by an AI agent — the kind of headline that makes you picture an AI picking victims and breaking in on its own. On Monday they walked it back: a human chose the target, set up the C2 infrastructure, and handed the AI stolen credentials. The AI ran the technical playbook — initial access, lateral movement, encryption, ransom note — but it was following a script, not freelancing.

I'd read this as: someone automated the grunt work of a known attack chain. That saves the attacker time on typing commands, but the reconnaissance and decision-making still sat with a human. We're nowhere near fully autonomous cybercrime. Sysdig hasn't named the victim, disclosed the ransom amount, or said which model was used. What's confirmed: an AI agent served as an automation engine in a real attack. That's notable, but it's not Skynet.

Source: TechCrunch · AI · techcrunch.com
Published: Jul 7, 2026, 07:56 UTC+8 · Jul 7, 2026