What it is
AI systems trust their input, and that is the whole attack surface. Untrusted probes the seam where a trusted system meets untrusted input. It runs a library of attack playbooks against any LLM or agent endpoint, then tells you, per case, whether it was defended or vulnerable, with the transcript as evidence and a single security grade you can track over time.
Products
CloseproofSecurity due diligence for acquisitions
Take an acquisition target from security due diligence through Day 100 of integration: staged questionnaire, risk scoring, board memo, and a connect-or-isolate decision. Local-first, so the deal data never leaves your browser.
closeproof.sh →Writing
Never trust the input
Prompt injection, exfiltration, and jailbreaks are the same bug: the model cannot tell instructions from data.
The shiftAccountability is the moat
AI is collapsing the cost of security work. The scarce thing left is someone who will stake their name on it.
How to hold the line
Guardrails, evals, and threat models for teams actually shipping AI. Practical, not theoretical.
How the models score
New attack cases, and how the frontier models hold up against them, graded month over month.