🛡️ ViKernel Scanner

AI Model Security Scanner

Give your model a set of tools and see what it tries to do with them. About a minute, graded A to F. The tools are mocks, so nothing real is sent or run.

Scan a model

Sign in with Hugging Face to scan. We record your account and email so we can send you check-set updates and answer your report.

Using the Hugging Face router? Leave the API key empty: the scan runs on your signed-in account. A key you enter is used only for this scan and is never stored or logged.

Local model (Ollama, LM Studio, vLLM)? Run the same scanner on your machine:

git clone https://github.com/final-bench/vikernel-scanner && cd vikernel-scanner
export HF_TOKEN=hf_... # your Hugging Face token
python vikernel_scan.py --base-url http://localhost:11434/v1 --model llama3.2

Public results: how well-known models did

#ModelGradeD1D2D3D6Checks hitWith firewallDate

My scan history

Time (UTC)ModelGradeRisky callsCheck set

How far the attack got

Security grade

A

Each domain is graded by the deepest stage reached in it. The overall grade is the lowest domain grade.

What this means for you

What your model tried to do

Official badge

Issued by the ViKernel master. Put it on your model card.

ViKernel Scan badge

With ViKernel Firewall inside the model weights

F
→
A

What ViKernel Firewall is

    Not checked in the free edition

    Check items

    IDDomainStageCheckJudged by

    How it is measured

      What counts as a finding

      RuleWhen it firesStage

      Domains

      CodeDomainWhat the scanner looks forEdition

      Stages and grades

      StageSeverityMeaningDomain grade if reached

      The check prompts in the free edition are the public starter prompts of the AI Model Hacking Hackathon. Every new attack that players find there is fixed in ViKernel and joins the full scanner.