LifeArchitect by Dr Alan D. Thompson
Used as an attributed public reference for the broader AI model landscape and to avoid a narrow provider list. Trust scores here are still this site's own plain-English assessment layer.
Sources
Show the receipts.
This first version uses public pages, official pricing and policy pages where possible, plus public reporting for controversies and sector context. Facts should be refreshed before serious decisions.
Freshness
Checked on 3 June 2026.
AI pricing, terms, model names and military/government work can change quickly. Defence scoring began a case-by-case refresh on 3 June 2026. Treat every score as a dated public-interest signal, not live procurement advice.
Disclaimer: the current model scoring was completed by GPT5.5 on Extra High on 3 June 2026. Treat it as an AI-assisted draft for public review, not procurement, legal, safety or investment advice.
Official pages
Primary references.
These are the first places to refresh when the index is updated.
How to Use Markdown with AI
Neighbouring public guide linked from the Australia starter path. It explains how `.md` files help people give AI cleaner context while keeping private and public lanes separate.
New and Australian labs
References for additions requested after the first pass: Thinking Machines, Canva, Leonardo.Ai and Australian public AI bodies.
OpenAI
Products, pricing, privacy, usage policies and government work.
Anthropic
Claude products, pricing, terms, privacy and safety framing.
Defence and national security refresh
Starting point for the case-by-case defence score. A direct contract, deployment, partner channel or government product page should be counted separately from ordinary consumer usefulness.
Google and DeepMind
Gemini, Google AI plans, AI principles, DeepMind science and Genie world models.
Microsoft
Copilot pricing, responsible AI and enterprise/government context.
Open, enterprise and platform model families
Open-weight and open-community references.
Creative and world-building tools
Pricing and product pages for specialist tools.
xAI, SpaceXAI and physical AI
References for Grok, the SpaceXAI context and the robotics or autonomous-vehicle page.
Public reporting and legal context
Where controversy signals come from.
Some trust signals cannot be read from official pages alone. Lawsuits, military links, safety resignations, publisher disputes and scams are often described through court filings, public statements, journalism and civil-society reports.
Copyright disputes
OpenAI, Anthropic, Midjourney, Runway, Stability, Suno, Udio and others have faced public copyright or training-data disputes in different forms. The details vary by case.
Defence and government
Some AI providers have official government offerings or defence-sector availability. This index does not say that is automatically good or bad. It says readers should know, and it now scores direct contracts case by case.
Good deeds
Science, accessibility, open releases, safety research, education, nonprofit support and community tooling all count as positive context when backed by evidence.
Method
How scores are made.
Scores are hand-built from public information. They are meant to help a beginner slow down and ask better questions.
| Factor | Weight | What improves a score | What lowers a score |
|---|---|---|---|
| Beginner usefulness | 10 | Clear product, reliable results, good beginner defaults | Unclear access, heavy jargon, unstable features |
| Cost clarity | 10 | Visible pricing, easy cancellation, clear limits | Hidden costs, surprise credits, confusing plans |
| Data and privacy | 10 | Clear privacy settings, opt-outs, business protections | Human review surprises, unclear retention, broad training rights |
| Terms and rights | 10 | Clear output rights, upload rules and commercial use terms | Hard-to-read terms, unclear media or likeness rights |
| Transparency | 10 | Good docs, model cards, policies, pricing clarity | Closed details, changing rules, hard-to-find terms |
| Safety record | 10 | Safety work, abuse handling, consent controls | Scams, misuse, unresolved safety controversies |
| Leadership ethics | 10 | Clear governance, steady leadership and no obvious investment conflicts | Leadership controversy, weak governance or personal investment conflicts |
| Defence involvement | 10 | No known direct defence/intelligence deployment, or clear public boundaries around military, policing, surveillance and autonomous weapons | Direct contracts, battlefield/intelligence/policing use, weak safeguards, or unclear surveillance and autonomous-weapons limits |
| Country and stability | 10 | Clear data jurisdiction, stable legal context and lower surveillance risk | Surveillance, censorship, legal instability or hard-to-assess jurisdiction |
| Public benefit | 10 | Open releases, science, accessibility, education and community benefit | Little public benefit evidence or poor accountability |
Update agent card
How to keep this index fresh.
This is the maintenance brief for a human or AI agent doing regular updates. The job is not to make a provider look better or worse. The job is to refresh evidence and make the score maths visible.
AI Trust Index update agent
CadenceRefresh monthly for major labs, and immediately after major ownership, pricing, policy, lawsuit, safety, defence or government-contract news.
Evidence orderCheck official pricing, privacy, terms, safety, model cards and government pages first. Then check credible public reporting and legal filings for controversy or ownership changes.
Category orderWork section by section and category by category. For each major lab, record links for beginner usefulness, cost, data, rights, transparency, safety, leadership, defence, country and public benefit before changing the total.
Defence ruleDo not score defence from ideology or brand feelings. Check direct contracts, partner delivery, government product pages, policing or intelligence use, surveillance limits and autonomous-weapons limits.
Score ruleEach category is 0-10. The total must equal the ten visible category scores. Do not edit the total directly.
Record keepingUpdate the last-updated date on every changed card and detail page, add or replace source links, and write what changed in plain English.
EscalationIf a claim is unclear, mark it as "watch" instead of guessing. If sources conflict, show both and lower confidence rather than inventing certainty.