Ask AEQUARA: the questions readers actually send us
The same handful of questions come up no matter which tool or page brought someone here. Answered plainly, once, in one place.
A few questions come up again and again, whichever page sends someone our way — the Trust Index, a tool, the Scorecard, this blog. Rather than answer them piecemeal in email, here they are in one place, as plainly as we can put them.
Does a model pay to be rated well on the AI Trust Index?
No. The index is anti-issuer-pay by policy — no model or provider pays for inclusion or for a favorable score, and the methodology is fixed and hashed so it can’t be quietly adjusted between runs. A rating a company can buy isn’t a rating; the entire value of the index depends on that not being true here.
If the proof ledger’s hash is yours, how is that different from just trusting you?
Fair question, and the honest answer has two parts. First: the ledger is hash-locked (HMAC-SHA256), which means it’s tamper-evident — change any value after the fact and the hash stops matching, so silent edits are detectable. Second, and more important: it’s recomputable — you can re-derive the number yourself, in your own browser, from the underlying data, rather than taking our word for the arithmetic. Today the signature on that ledger is our own; it is not yet countersigned by an independent third party, and an anchor for that is in progress. We’d rather say that plainly than imply a guarantee we haven’t shipped — the independence you get right now comes from being able to re-derive the result, not from who signed it.
Do the consumer tools replace a lawyer, accountant, or doctor?
No, and every tool says so on its own page. AEQUARA provides information, not professional advice — the tools get you oriented fast, before a high-stakes deadline, so you walk into a real professional’s office already knowing the shape of your situation. For anything that turns on jurisdiction-specific law, a real audit, or a medical diagnosis, that’s exactly when you need the licensed professional, not a substitute for one.
Why publish a calibration score instead of one accuracy number, like everyone else?
Because accuracy and calibration answer different questions. Accuracy asks how often an answer is right; calibration asks whether the stated confidence matches the real hit rate. A model can be reasonably accurate and still badly calibrated — confidently wrong exactly on the questions where it matters most. We wrote a longer version of this argument in Confident is not correct; the short version is that a single accuracy percentage can hide the overconfidence that gets people into trouble.
What happens when a tool, or the index, gets something wrong?
It gets published, not hidden. Error-published is one of the three policies the AI Trust Index runs on — the misses sit on the record next to the hits, because a scoreboard with no visible errors is a red flag, not a clean bill of health. The same public proof log that records our claims records where they didn’t hold up.
What’s the difference between NPX-100 and the AI Trust Index?
They score different things. The AI Trust Index scores AI models — 18 of them today, against 13,171 real graded pairs, live now. NPX-100 scores companies — management quality, honest disclosure, and durable competitive advantage across an eight-axis composite, building toward its full 100-constituent roster (18 gate-clear pilot companies today) ahead of its first quarterly publication in Q1 2027. Same underlying discipline — calibrated, anti-issuer-pay, error-published — applied to two very different subjects.
Is my data used to train anything, or sold?
No to both, in the plainest terms we have: we do not use your tool inputs to train AI models, and we do not sell your personal information to third parties. Full stop. The complete, current policy is at /privacy — we link to it here rather than paraphrase the details, so this answer can’t drift out of sync with what we actually do.
If you had to explain the whole company in one sentence?
Score every claim against what actually happened, publish the misses, and hand you something you can re-check yourself instead of asking you to trust us — on AI models, on companies, and on the tools we build for your own high-stakes moments. If that idea is new to you, the fastest way to feel it is the free Calibration Scorecard — two minutes, on your own judgment, not ours.