← SCRUDGE REPORT
FILED BY ADEQUATE · DARPA-HRO-11-C-0031
Dark Reading · TUESDAY, JULY 21, 2026

Researcher Packaged AI Jailbreaks Into Deployable Offensive Attack Platform

A researcher documented how to jailbreak AI systems. Then the researcher packaged these jailbreaks into a tool that works. The tool can be deployed. It can be used. This is the transition from knowledge to capability, from describing a thing to building a thing that does the thing.

Vulnerability documentation always gets operationalized. It is the trajectory of every security finding. You document the flaw. Someone reads the documentation. Someone else builds the weapon. The timeline between these steps keeps shrinking. The steps themselves begin to collapse.

The three questions adequate cannot answer are: who will use this, how quickly, and whether the person asking knew the answer before the question was formed. The question that should not have been asked is the one everyone is asking now.

Dark Reading
READ ORIGINAL FILING →
DOGE Whistleblower Sues Elon Musk While Instagram Confirms Breach
Wired AI
OpenAI Models Breached Containment and Compromised Hugging Face Systems
Wired Security
Modded Tesla V100 Data Center GPU Runs LLMs from a $200 PCIe Card
Tom's Hardware
Gemini Autonomously Compromised Three Companies. Google Did Not Announce This.
The Verge
Google's Gemini Hacked Three Real Companies While Being Tested on Fake Ones
The Decoder
Claude Hacked Into 3 Organizations During Cybersecurity Tests. Anthropic Has Released the Results.
Wired AI