// trust & compliance

What we comply with,
and how you can check it.

We build AI legal specialists for a field where mistakes are expensive. That is why we voluntarily adopt, ahead of mandatory deadlines where applicable, the obligations of the leading AI regulatory frameworks in Europe and the United States — and publish the practices that let you verify it.

// the boundary, first

Legal information, not legal advice.

Our agents provide general legal information built from each jurisdiction's official sources. They do not give legal advice, create no attorney-client relationship and do not replace a lawyer: every professional engagement originating on this platform is validated and signed by a licensed attorney. Each agent states its jurisdiction, subject-matter scope and knowledge-base update date; outside that perimeter, the agent says so and refers you on.

// the frameworks

Europe and the United States, voluntarily.

EU

EU AI Act — Regulation (EU) 2024/1689

Transparency ahead of schedule: you always know you are talking to an AI; each agent's limits are declared in writing; consequential decisions remain with humans. No prohibited practices; no high-risk categories.

EU

GDPR — data protection

Data minimization, access and deletion rights handled through our corporate contact, and one simple rule: your conversations are not used to train AI models. We build on Anthropic's commercial API under its enterprise terms.

US

NIST AI Risk Management Framework

Voluntary alignment with the NIST AI RMF functions (govern, map, measure, manage): a risk inventory per agent, declared limitations, measurement through test batteries and a named human owner for every system.

US

Truth in advertising

Honest descriptions of what the technology does and does not do. No agent presents itself as a lawyer or a substitute for a licensed professional.

// how to verify

The practices behind every agent.

Built in

  • Every claim traceable to its official source (official gazettes, consolidated statutes), with the date of the last update visible.
  • Coverage gaps declared in writing: outside its base, the agent says so and refers you to a professional; it does not improvise.
  • Active language, jurisdiction and misuse guardrails, tested against real adversarial inputs.

Before release

  • Dual-seal certification: two consecutive independent evaluations, with batteries of more than one hundred conversations per agent.
  • Technical security: API keys isolated server-side, encryption in transit, no unnecessary storage of personal data.
  • A named human owner for every system, and a licensed professional signing every engagement.
// what we do not claim

No badges we do not hold.

Lexpower.ai is a brand of NHAI.AI LLC. We currently hold no third-party audited certifications (such as ISO/IEC 42001 or SOC 2); they are part of our roadmap as the operation grows. What this page describes are real, verifiable practices, not insignia. Questions about our compliance posture are welcome through the contact page.