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.
Status: preprints, submitted to AIS26; peer review in progress. Not yet a published proceedings paper — stated precisely. Our page for the summit →
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.
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.
The point of the design is that a third party can confirm every claim without trusting our servers.
A real ML-DSA-65-signed credential + RFC-3161 timestamp, checked offline against a published did:web key — no issuer server involved.
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.