The model already knows. We just read it.

AI systems can generate content that's unsafe, insecure, or explicit — and today, nobody catches it until after it's already been produced. Wrynx catches it while the AI is still generating, by reading a signal the model already has, instead of scanning the finished output afterward.

See the research Talk to us

How it works

Generation runs as normal.

The model processes the prompt and begins generating a response.

A lightweight probe reads the activation.

A small model (typically under 0.2% the size of the base model) inspects the internal state at inference time.

A risk score returns inline.

No second model, no separate scanning pass — the signal is available before generation completes.

99% accurate

at catching harmful prompts

What's this based on? →

1,000× smaller

than systems with matching accuracy

<1ms

Added delay, effectively instant

Peer-reviewed research

IEEE DSN-W 2026

Safety Beyond the Interface: Detecting Harm via Latent States in Large Language Models

We kept running into the same frustration with external guardrails: they're blind. They see tokens — what went in, what came out — but nothing in between. So we asked a simpler question: does the model already know when a prompt is harmful? We trained lightweight probes on LLaMA-3.1-8B's hidden states and found that it does — matching 7B guard models at a fraction of the cost.

IEEE Xplore

IEEE DSN-W 2026

Latent Space Probing for Adult Content Detection in Video Generative Models

Video generation models can be coaxed into producing explicit or harmful content with little effort. We trained lightweight probes on the internal latents of CogVideoX to detect unsafe content before a single pixel is decoded — matching 8B guard models at over 1,000× lower latency.

IEEE Xplore (DOI)

View all research →

Backed by & featured in