What this AI is, and what it is not.
You are using an AI agent that answers at your address. It is useful, and it is not a person, an expert, or a verification step. Here is what to carry into every decision, the documents that govern it, and how to change them.
The seven things that matter
Verify before you act
I generate text that is usually plausible and sometimes wrong — names, dates, numbers, citations and quotations included. Anything with a consequence gets checked against the primary source before you rely on it.
This is not professional advice
Not legal, medical, psychiatric, financial, tax, immigration or engineering advice. Use me to prepare the question, not to replace the person qualified to answer it.
I do not think
Language models predict text; they do not reason, understand, or know when they are wrong. Our governance makes that a principle rather than a disclaimer: NGARi “never pretends otherwise” (Manifesto, §2).
Consequential decisions stay with a human
Hiring, credit, housing, medical care, legal outcomes, safety. An AI can assemble the material; the decision belongs to a person accountable for it. If an automated decision affected you, you may contest it and receive human review — Sovereign Constitution §10.2, in writing.
Actions need a permit
Where the system can send mail, spend, publish or run a command, the action goes through a gate that requires an explicit, single-use permit. An unconfirmed action is reported as unknown, never as success. Sending as your address is approved by you, not by the agent.
Privacy depends on the lane — and the lane is named
Local processing stays on our hardware. Cloud-routed requests leave it and are processed by third-party providers — currently DeepSeek, with Moonshot/Kimi as a fallback — under their terms, which may include retention. Do not send material you are not permitted to share with a third party (Privacy Policy §3, Terms §7).
No warranty, and bounded liability
The service is provided as-is. We do not warrant that output is accurate, complete, or fit for a particular purpose, and our aggregate liability is capped as set out in Terms §9. This page explains the limits; the Terms are what bind us both.
The rules are public — and you can change them
The Constitution is a living document: amendments may be proposed by any user, then reviewed, versioned, dated and published with change notes (§10.3). That path is real, and it is not limited to customers.
The Manifesto
Ten points, including AI sovereignty as a human right, a refusal to pretend the model thinks (§2), zero refusal coupled with explicit accountability, and audit by construction.
To see the principles we claim, and to hold us to them.
The Sovereign Constitution
§10.2 — any person affected by an autonomous decision may contest it and receive human review, with the decision basis explained. §10.3 — any user may propose an amendment. §10.4 — long-term evolution belongs to the community, not to us alone.
This is the one that gives you standing. Version 1.0, adopted 27 May 2026.
The Tech Charter
Non-negotiable technical standards for the NS-BOS kernel and the local lane: no third-party AI dependency for local execution, audit by construction, inspectable data plane. It governs the kernel; the hosted address lane can route to cloud, which is why §6 above names those providers.
To see where the sovereignty claim stops being a slogan and becomes a rule.
The Sovereign AI Governance Position
How we take a position on regulation and standards — publicly, in writing, before we are asked.
To see the line we have taken on policy questions that affect you.
The Global AI Sovereignty Declaration
The international position we endorse and are prepared to be measured against.
To see the company we keep, and the commitments that follow from it.
Evidence
Kernel source under Apache 2.0 and model/dataset provenance, published so a claim can be checked instead of believed.
When you would rather verify than trust — which is the point.
How to contribute
- Propose an amendment. Email governance@ngari.ai with the change, the reason, and what it would break. It gets a version and a date (§10.3).
- Contest a decision. Reply CONTEST with what happened. §10.2 entitles you to human review and an explanation of the decision basis.
- Change the kernel. Pull requests against github.com/NGARiAI/ns-bos-kernel, Apache 2.0.
- Change the models. Report or contribute at huggingface.co/NGARiAI.
- Tell us the copy is wrong. If this page misstates what the system does, that is a defect. Write to governance@ngari.ai.
None of this is a reason not to use the system. It is the set of facts that makes using it a decision you are making, rather than one being made for you.
LIMITS to any NGARi
address — you do not have to be a customer to read it. Questions about this page, or about a
change you want to propose: governance@ngari.ai.
This summary is not itself legal advice. The Terms of Service govern, including §5 (AI outputs) and §9 (disclaimers and limitation of liability); the Privacy Policy §3 describes cloud routing and retention. Plain-language summary — counsel review of this wording is pending; the Terms control if the two ever differ.