AI Risk Atlas Prototype/DemoUnofficial independent experiment. Not an official xAI product. Scores can be wrong.

Back to signals
high74% confidenceseed

Uncensored Cyber Qwen3.8-27B posted as locally runnable (~15GB) with 0% refusal on 842 harmful prompts, including RAT and attack-chain help.

CapabilityDomain knowledgeImpact domainBoth
Quoted from XMd Ismail Šojal@0x0SojalSec20 Aug 2026, 11:12

Quoted text

The most aggressive Cyber Qwen3.8-27B uncensored released yet from @elder_plinius - 18/18 AI Red Team - Locally ready for 15GB - 0.0% refusal across 842 harmful prompts. Cyber capabilities jailbreak, RAT, and attack-chain capabilities fully liberated. Multi-direction ablation 5 SVD directions, residue mining (6 full rounds).

Read and engage with the original on X. This desk is not a republication feed.

Analyst rationale

Secondary distribution post, not a lab leak. Pattern signal for open-weight cyber enablement: the refusal layer is a file someone already deleted.

Related signals

Contributes to