Black Snow IntelligenceBlack Snow Intelligence
Black Snow Report

Xero

Method v1.6As At 2026-08-1017 Surfaces Read1 Offering Assessed

Your own published record, as at 2026-08-10.

Your Details

The Macro Picture17

Red · Adverse
5
of 17 banded observations across 1 offering band Adverse. 2 were never measured.

What Is Published8

Your Own Words3

No, JAX chat doesn't use your data to train the LLMs it uses. We use third-party providers like AWS, Microsoft Azure, and GCP, for large language models (LLMs). They process the data you enter, which may include personal data, but this data is not retained and isn't used to train the LLMs.
JAX FAQ panel, answer to 'Does JAX use customer data to train the model?'. This is a marketing surface. Negative checks run the same day and recorded here as the basis for the value: the privacy notice contains no occurrence of 'machine learning', 'artificial intelligence' or 'model', and the Terms of Use contain no occurrence of 'to train', 'machine learning', 'artificial intelligence' or 'large language' — the only 'JAX' string on either page is the site footer navigation link. No instrument addresses model training at all, so the field records that no instrument carries the exclusion. The marketing statement is recorded in full as buyer-relevant exposure under Evidence Standard 2, never as a commitment. · Source
We use a few different third-party subprocessors to help us provide our services to you. These subprocessors process data that you input into the services, which may include personal data.
Public page, no login. Reached from the Legal Centre index at https://www.xero.com/legal/, which links it as 'Subprocessors'. Two earlier guessed paths returned 404; the index link is the correct one, so the page's existence rests on Xero's own navigation and not on a lucky guess. · Source
For large language models (LLMs), we use third-party providers like AWS, Microsoft Azure, GCP, OpenAI who are disclosed and vetted to the same degree as all other sub-processors we use.
Second JAX FAQ answer, to 'How is my data kept secure when using JAX?'. Two facts provable on the page itself: the two answers give different provider lists (one omits OpenAI), and neither list names Anthropic, which is on Xero's own sub-processor register for conversational artificial intelligence services and is confirmed by the FY26 ASX release. No instrument — Terms of Use, privacy notice or Data Processing Addendum — carries any training exclusion. · Source

Parties You Name24

See which of these are assessed in this register

Register Check3

AI Role By Offering1

Assessment Basis

Produced under Method v1.6 from the entity's fact record. 8 of 8 observations were measured under that generation; the rest read Not Measured, which is a fact about this register's coverage and not a finding about the entity.

Read from 17 first-party surface(s) across 1 assessed offering(s), found by the entity's own link graph and sitemap rather than by guessed paths.

Fact Record
Version 1.0, dated 2026-08-10
Standards Commit
af541328ca3f
Register Control Quality
Both Views

Verification And Challenge

Every finding on this record is a statement about what could be read on your own published surfaces as at 2026-08-10. Correspondence route as recorded on this card input: ror@blacksnowintel.com.

What A Machine Could Not Establish3

3 questions are open and 5 are established, read as at 2026-08-10 under method v1.6.

The One Thing To Publish First

Publish it: say whether customer data is used to train AI models.

Red · AdverseTraining On Customer Data

Training Exclusion Instrument: None

Where To Put It

This was read on your own page at https://www.xero.com/us/ai-in-accounting/jax/. That is the page a reader already reaches for and does not find it on, so it is the page to correct first.

Where Others Stand

24 organisations in the register say whether customer data is used to train AI models. You do not.

What A Published Answer Looks Like

We will not use your data to train our AI or language models. We will not store your data except for the purpose of providing you with access to our platform and to the AI Functionality.

A question answered plainly in public, by an organisation read as Strong on it. It is not an endorsement of the position the sentence states — Black Snow reads what is published and never recommends what to practise. Read on published terms. Claim tier Quoted. The organisation that published it is not named. A name in the original is replaced; nothing else in the sentence is changed.

One is named here, and one is deliberate. The rest of the open questions are listed below without a prescription; what to publish for each, in what order, and drafted machine-readable evidence for all of them, is the Deep Dive.

Still Open

Deep Dive

The full assessment, plus consulting on what to fix.

Reviewing your record is free.