One trace from a high-risk model, mapped to a specific EU AI Act obligation, verifiable in roughly sixty seconds.

Warrant is regulator-grade evidence infrastructure for AI agents in regulated industries: drop an agent's execution trace, get a record mapped to a specific EU AI Act obligation, independently verifiable without contacting Warrant, in roughly 60 seconds.

Drop a trace JSON or pick a sample. Warrant reads the trace, scopes the regimes, weighs each action, and emits a PDF evidence package with article-level citations — a record mapped to a specific EU AI Act obligation, independently verifiable without contacting Warrant. The artefact below is what you would file with a regulator.

The sample lending trace maps to EU AI Act Article 12 — Record-keeping and Article 13 — Transparency to Deployers. Regimes the classifier puts in scope but for which the obligation corpus carries no table — NYDFS 23 NYCRR Part 500 among them — are reported on the package as classified and in scope, not evaluated.

// v0.4 public beta · evidence of warrant's analysis, not legal advice. the PDF is independently verifiable without contacting Warrant the moment it downloads.