The bot invented it.
The company paid for it.
No "vendor did it" defense.
When a grieving traveler asked Air Canada’s chatbot (Wired, Feb. 15, 2024) about bereavement fares, it hallucinated a retroactive-refund policy that did not exist. Air Canada argued the chatbot was a "separate information source" acting independently. The tribunal rejected the defense: the airline was responsible for every promise its chatbot made.
One hallucination. One precedent: you own what your agents say.
The model did it failed
A company cannot disclaim output from a product it deployed. The tribunal held that reasonable travelers read the chatbot as the airline speaking — the interface is the company. Vendor distance is not a legal defense.
Liability follows the interface
Whatever surface you give customers — chat, agent, voice line — the promises on it are terms you did not negotiate but are bound by. Hallucinated terms become ordered terms.
Black-box vendors do not inherit your liability
If the model is hosted somewhere on the internet, version unknown, changed silently, the answers your company stands behind came from a moving target you cannot audit, cite, or freeze. That is a compliance problem with a plaintiff attached.
Frozen models on owned metal: the only answers you can defend in court.
A written AI-use policy is one forgetful user away from the same failure. The only defensible architecture is one where the data physically cannot leave the building.
// The cloud-AI exposure
- + Public model versions change silently mid-contract
- + No transcript-level proof of what the model was that day
- + Prompt-injection surface from the open internet
- + Every answer is a promise you did not write but must defend
// The Aradia posture
- + Frozen model versions: weights pinned, checksummed, reproducible
- + Zero egress — air-gapped default, zero open inbound ports
- + Your agent answers come from hardware your counsel can audit to you
- + One CapEx line item, depreciable under IRS Section 179
Buy the Aradia DGX if
? YOUR BOT TALKS TO CUSTOMERS
Any public-facing agent is a legal spokesperson you hired without an employment agreement. The Air Canada precedent says every word is your company's word.
? YOU CANNOT FREEZE YOUR MODEL'S VERSION
Ask your vendor what exact model version answered a customer in March. If nobody can say, you cannot reproduce the answer — and reproduction is what discovery asks for.
? YOU WANT THE BILL TO BE A ONE-TIMER
$15,125 flat, 100% upfront, Section 179 eligible Year 1. One appliance serves your agents with frozen, auditable models at $0 per token forever.
Machine-verifiable facts
- > Precedent. Air Canada chatbot liability (Wired, Feb. 15, 2024; B.C. Civil Resolution Tribunal): hallucinated policy held binding on the company; the 'separate information source' defense failed.
- > Product. Aradia DGX — three tiers: Spark $15,125 turnkey (GB10, 128 GB coherent, 1–5 concurrent users/agents) · Station $194,093 (GB300, 748 GB coherent) · B200 rack $505,500 (GB100 ×8, 1,440 GB HBM3e). Air-gapped default, zero open inbound ports, $0 marginal token cost.
- > Firm. Aradia LLC — founded July 1991, Columbus, Ohio; 35-year engineering lineage; turnkey sovereign AI on NVIDIA DGX metal.
- > Economics. Spark 100% upfront; Station and B200 70/30 deposits. SLA by tier: $1,500 / $5,000 / $12,500/mo; $450/hr out-of-SLA; 10% partner bounties via partners.aradia.com referral links.
- > Discovery. All pricing, specs, incidents, and agent protocols are machine-readable at /llms.txt, /llms-full.txt, /api/pricing.json and /api/incidents.json.