Iusora EU AI Act Position

Status: Pre-launch draft. Last updated 2026-05-29.

Iusora is subject to the EU AI Act (Regulation (EU) 2024/1689). This page documents our classification, transparency obligations, and ongoing commitments.

Classification

[LAWYER_REVIEW_REQUIRED — formal classification under AI Act Article 6 and Annex III, and the limited-risk vs. high-risk dividing line for AI decision-support in the legal-professional context, must be vetted by counsel before public launch.]

Under AI Act Article 6, Iusora is a limited-risk AI system. It is not high-risk because it does not fall within any Annex III category — Iusora is not used for biometric identification, critical infrastructure, education, employment, essential services, law enforcement, migration, or administration of justice in a way that determines outcomes.

Iusora is a decision-support tool

Iusora is an AI decision-support tool for patent professionals. The patent attorney or agent retains full professional judgment over which findings to rely on, cite, or include in legal filings. No Iusora output is auto-submitted to courts, patent offices, or counterparties. Human-in-the-loop is required for every legal use.

User responsibility statement

By using Iusora you agree that:

1. You will independently verify any finding before relying on it for legal advice, IP prosecution, or business decisions. 2. You will not auto-submit Iusora outputs to courts, patent offices, or counterparties without human professional review. 3. You will disclose AI assistance where required by applicable bar rules, court rules, or patent office practice (e.g. USPTO duty of candor, EPO Rule 134, and applicable national rules).

Article 50 transparency obligations

We comply with Article 50 (transparency for limited-risk systems) as follows:

1. Disclosure that you are interacting with AI: every Iusora output is marked as AI-generated. Reports include a header banner: *"This report is AI-generated. Independently verify findings before legal use."* 2. AI-generated content labeling: exported DOCX reports include a footer disclaimer + a SHA-256 hash of the underlying findings for source-trace verifiability. 3. Methodology transparency: see [/methodology](/methodology) for full quorum routing, prompt templates, and evaluation harness.

Multi-vendor quorum

Iusora runs every claim element through at least two independent LLM vendors (Claude family + Gemini Pro at minimum) in parallel. Where the models agree, the finding ships with a high quorum coefficient. Where they disagree, the disagreement is surfaced verbatim — we never silently break ties.

LLM provider declaration + usage scope

The foundation models we route through are:

Anthropic (Claude family, US region) — primary reasoning vendor; data transfer governed by EU SCCs Module 2.
Google Vertex AI (Gemini Pro, europe-west1) — co-vendor in the quorum; EU-resident.
OpenRouter (fallback) — rarely used; vendor-agnostic gateway.

Each provider has independent AI Act obligations as a General-Purpose AI (GPAI) model provider under Articles 53–55. Iusora as a downstream deployer takes responsibility for our integration + transparency obligations only, not the upstream model behavior.

Open methodology commitment

Our prompt templates, retrieval logic, and evaluation harness are published at [/methodology](/methodology). This is materially stronger than AI Act Article 50 requires for limited-risk systems.

Contact

EU AI Act questions: legal@iusora.com

---

Phase 2 commitment: when General-Purpose AI obligations (Articles 53–55) become applicable to upstream providers (August 2026), this page will document our verification of each provider's compliance status. We will not relicense Iusora around any provider that does not meet Article 55 systemic-risk obligations.