Paste the list of datasets behind your AI models. See which ones you cannot account for.
AI Training Data Provenance Checker reads your dataset list line by line for provenance, labelling and lineage gaps against EU AI Act Article 10 and ISO/IEC 42001. Paste the list of datasets behind your AI models, the list and never the data. Every dataset is checked for where it came from, the terms it came under, personal and special category data, who labelled it, which version it is, and whether the same data sits in train and test, each finding with the EU AI Act Article 10 or ISO/IEC 42001 clause behind it.
Paste the list of your datasets, never the data. One line per dataset: name, source, terms, personal data, who labelled it, version. The list is read in your browser and nothing is stored until you save. A cell that looks like record content, such as an email address or an identifier number, is removed before anything is shown; review the list before you save it.
It reads what your list says about each dataset. It does not open or scan the data, it does not match your files against other collections, and it gives no score.
Every finding names the column that raised it and the clause behind it, and says what it could not determine: terms not recorded, source not recorded.
It never says a dataset is lawful, licensed, cleared or compliant.
- D-01trainClaims notes 2019-2024internal claims system; internal; labelled in house; snapshot 2024-12-31
- D-04trainMotor forum postsweb scrape; [no terms recorded]; labelled by a crowd; [no version]
- D-08trainImage set for vehicle damagevendor; vendor licence; labelled by a vendor; [no version]; [provider not named]
- D-30trainVehicle images from assessors[origin not recorded]; [no terms recorded]; labelled by a crowd; v3
- D-10testEval set Q3copy of claims notes, Q3 cut; internal; labelled in house; eval-q3
and 6 more datasets in this chain
32 datasets, 7 with incomplete provenance records.
12 finding types, 2 declared potential overlaps, 31 of 32 placed.

Paste the list as it sits in the spreadsheet
One dataset per row: dataset | source | terms at least, or with any of feeds, split, personal data, sensitive categories, lawful basis, labelled by, label check, version, collected, provider and bias examined. Each line is placed on one of 42 published source classes and 26 terms families; a line that fits none is marked unplaced and never guessed.
Read the chain for each model
Every model with the datasets that feed it in split order, each written origin; terms; labelled by; version. A link your list does not record is printed in square brackets and marked, and the chain breaks there. Declared potential overlaps between train and test are drawn as pairs.
Take the datasheet pack to the dataset owners
Twelve findings in fixed order, each with the datasets it names, the clause from ISO/IEC 42001, the NIST AI RMF, the EU AI Act, GDPR or UK GDPR, and the question to put to the dataset owner or to legal.
Why a list, and not a scan
Data lineage software connects to your systems and traces tables. The head of data governance at a 900-person company is asked a simpler question first: which datasets trained this model, where did each come from, and under what terms. The answer sits in a spreadsheet kept by whoever ran the last project. Before anyone buys a lineage tool, somebody has to show which lines of that spreadsheet have no origin, no terms, no version or no check on the labels, and which clause asks for each. That is what this reads, in your browser, from the list as it stands.
Training data provenance: the questions data governance leads ask
What is AI Training Data Provenance Checker?
It reads your dataset list line by line for provenance, labelling and lineage gaps against EU AI Act Article 10 and ISO/IEC 42001. Every dataset is checked for where it came from, the terms it came under, personal and special category data, who labelled it, which version it is, and whether the same data sits in train and test.
Do I paste the data or a list?
You paste the list of your datasets, never the data. One line per dataset: name, source, terms, personal data, who labelled it, and version.
Which regimes does it check against?
Each finding carries the clause from ISO/IEC 42001, the NIST AI Risk Management Framework, the EU AI Act, GDPR or UK GDPR, read clause by clause from the published text. The EU copyright directive's text and data mining provision, US fair use and the EU AI Office's training content summary template are named, not quoted.
Does it tell me whether a dataset is lawful or licensed?
No. It never says a dataset is lawful, licensed, cleared or compliant. It reads what your list says about each dataset and raises the gaps as questions for the dataset owner or for legal.
What findings does it raise?
It raises twelve findings in a fixed order, each naming the datasets it concerns and the clause behind it. Every finding names the column that raised it and says what it could not determine, such as terms not recorded or source not recorded.
Does it scan or open my data?
No. It reads what your list says about each dataset; it does not open or scan the data and does not match your files against other collections. The list is read in your browser and nothing is stored until you save.
Does it score my list?
No. It gives no score. Each line is placed on one of 42 published source classes and 26 terms families, and a line that fits none is marked unplaced and never guessed.
Is AI Training Data Provenance Checker free?
Yes, eight lines of your own list are free with no account. The datasheet pack, the findings CSV, the board summary, the register CSV and saving the list are on Solo.
What happens to cells that look like record content?
A cell that looks like record content, such as an email address or an identifier number, is removed before anything is shown. You review the list before you save it.
What is the datasheet pack?
On Solo, the datasheet pack takes the findings to the dataset owners, each with the datasets it names, the clause behind it and the question to put to the dataset owner or to legal. The findings CSV, board summary and register CSV export the same work.
What does scope mean for each regime?
Scope follows each regime's own trigger. The EU AI Act applies where you are placed on the EU market, put into service in the EU, or your output is used in the EU, while GDPR and UK GDPR follow establishment or processing personal data of people in the EU or UK. You also state whether you provide a general-purpose model and your high-risk use, if any.
What is the EU training data summary?
The checker reads which lines of your list feed the EU training data summary, and the datasheet template shows the columns a dataset list needs. The EU AI Office's training content summary template is named, not quoted.
What this does, and what a spreadsheet, a lawyer or an AI assistant does better
A spreadsheet is better for holding the dataset list, adding the fields your own governance uses, and keeping the record without a tool in the loop. A lawyer or data protection consultant is better on whether a dataset is lawful, licensed or cleared, which this tool never decides. A general AI assistant is better for drafting the model card, the datasheet or the explanation a reviewer wants in prose. What this does that they do not is read the pasted list line by line, place each dataset on a published source class and terms family, mark the links the list does not record, and raise each gap as a question against the EU AI Act Article 10 or ISO/IEC 42001 clause it cites; it reads the list and never the data, and it never says a dataset is lawful, licensed, cleared or compliant.
Primary source: EU AI Act