Demonstrated, not asserted

Research & independent-verifiable evidence

Our claims are peer-review-format, measured on public corpora, and reproducible by a third party. We publish our failure modes as first-class results. Below: the preprints, the numbers, and how to check them yourself.

Preprints

Two working papers — AI Summit Styria 2026 (IT-Security)

Status: preprints, submitted to AIS26; peer review in progress. Not yet a published proceedings paper — stated precisely. Our page for the summit →

Preprint · in reviewIEEE formatPost-quantumEU AI Act

A Sovereign, Reproducible Trust Layer for LLM Systems

Two-Number Prompt-Injection Detection with Post-Quantum, Browser-Verifiable Audit Receipts

A sovereign, CPU-first LLM trust layer. AEGIS reports injection detection and content safety as two separate numbers, never fused into one; LEDGER/POAW emits FIPS-204 ML-DSA-65-signed, RFC-3161-timestamped, browser-verifiable receipts mapped to EU AI Act Art. 12/Annex IV. Includes a deception-indistinguishability measurement (four attacker models + a same-population control) and candidly disclosed gaps.

AEGISrates under public rerun
3309-bytereal ML-DSA-65 sigs
96crypto assertions
Download PDF (5 pp)
Preprint · in reviewIEEE formatFairnessMultilingual

English-Tuned by Default

Measuring and Mitigating Multilingual Over-Blocking in a Sovereign LLM Injection Firewall

A measured fairness study: an English-tuned guardrail over-blocks benign non-English users several times more often than English ones. An opt-in stopword-ratio dampener reduces the over-blocking without changing the English side — a non-discrimination dimension the AI Act cares about, measured and mitigated honestly.

Over-blockingnon-English benign text
Dampeneropt-in mitigation
n=46,440public corpus
Download PDF (4 pp)
Don't trust us — check it

Independently verifiable

The point of the design is that a third party can confirm every claim without trusting our servers.

Browser-verify a receipt

A real ML-DSA-65-signed credential + RFC-3161 timestamp, checked offline against a published did:web key — no issuer server involved.

Open the verifier →

AEGIS numbers: rerun pending

The preprints report rates from our own runs. The public result file and the preprints do not yet agree, so we show no AEGIS rate on this site until a reproducible public rerun is published. The preprints stay available as the dated record of what we submitted.

▶ 90-second POAW explainer

Honest scope. Two subsystems are proven and measured (AEGIS injection guardrail; POAW audit substrate); other cascade layers are research and labelled as such. We disclose gaps in the same breath as strengths — narrative-jailbreak detection is capped, key custody is software-only today (hardware roadmap), and "mapping to EU AI Act duties" is not a claim of conformity. Core methods are patent pending (provisionals filed 03-04/2026; non-provisional conversion due 2027). This restraint is deliberate — it is why the evidence is trustworthy.