DataPacts publishes on one rule: what a page includes, how it orders the entries and how it describes them never depends on who pays. This policy sets out that rule and the standards that follow from it: inclusion, tone, balance on decision pages, the limits of the information given, corrections and authorship.
The policy applies to every page of the reference on AI training data licensing, for sellers and buyers alike. It describes what readers can expect from the site and what they can hold it to, whether they hold data to license, work for a buyer or advise either side.
Fact-check note: every figure on the site carries a source and a date, as set out in how we source, date and check every figure.
Inclusion and order never depend on fees
Every buyer DataPacts finds is covered, whether or not it pays referral fees, and no buyer can pay to be listed, placed higher or described differently. Coverage is decided by whether a buyer or program exists and can be sourced to a primary document, not by any commercial relationship. A buyer that offers a referral fee is described in the same fields as one that offers none, and a buyer that declines to comment is still covered from its public statements.
Lists and tables are ordered alphabetically or by a stated neutral criterion such as date or data type, and never by payout. The site never applies these labels to a buyer:
- partner
- recommended
- verified
- trusted
- top
- preferred
A reader who wants to compare programs finds every buyer and program, compared without fees, with the same fields and the same check dates for each entry.
Neutral tone without urgency or promises
Pages are written in a neutral, encyclopedic tone, without hype, invented urgency or calls to sell. No page urges a reader to sell at once or to hurry, and no page or form promises a price, a buyer, a response time or an outcome.
A buyer's figures appear as that buyer's own claims with the date they were checked. Tone therefore never adopts a buyer's marketing language, and a page never presents a payout range as a result a seller can expect. Where a buyer's own pages give two different figures for the same thing, both are shown with their dates and neither is averaged or chosen.
Benefits, risks, downsides and alternatives on decision pages
Every page that helps a reader decide whether to sell or license data lists the benefits, the risks, the downsides and the alternatives, including not selling or licensing at all. This covers the seller process, where a reader can learn how to sell or license company data to AI companies, and also the guides, seller types, data types, valuation, contract and insolvency pages and the buyer profiles.
The table below compares the four labeled lists that each decision page carries (as of 2026-11-09; source: DataPacts editorial policy).
| Labeled list | What it contains | Example item type |
|---|---|---|
| Benefits | what a holder can gain from licensing or selling | revenue from records that are otherwise dormant |
| Risks | what can go wrong, with dated sources where an incident exists | rights that cannot be recalled once a model is trained |
| Downsides | costs and limits even when nothing goes wrong | time for rights audit, consent and de-identification |
| Alternatives | other routes, always including not selling or licensing at all | keeping the data; licensing a narrower set; not selling |
Not selling or licensing is a valid outcome and is always shown as one. A page that listed only benefits would not meet this standard, and neither would a page that listed only risks. Where a risk rests on an incident, the page dates the source.
Information, not legal or tax advice
DataPacts explains law and tax as information and does not give legal or tax advice for any reader's situation. A reader whose data, contracts or tax position raises a legal question needs a qualified adviser in the relevant jurisdiction.
Law pages, including the overview of AI training data law by jurisdiction, carry one plain sentence saying that the page is information and not legal advice. Every legal statement names the instrument, the article or section, and a date, such as the date a rule is in force or a case was decided.
Tax and accounting content follows the same rule. It describes how a rule treats a type of transaction and does not tell any reader how to report their own.
Corrections: what is corrected and how it shows
Factual errors are corrected on the page once a primary source confirms the correct value, and the corrected value carries a new check date. A correction replaces the wrong value in place, so a reader never has to compare versions to find the right figure.
Corrections reach the editorial team at editorial@datapacts.com. A public corrections log is planned, and until it is published the changes appear on the pages themselves, with the corrections log to follow as the central record. Each correction states what was wrong, what the page now says and the date of the check. No response time is promised for any request.
How buyers and other named parties can request a factual correction
A buyer, seller, court party or publisher named on DataPacts can request a factual correction by email, with a primary source that shows the correct value. A request that carries a primary source is checked against it, and a request without one cannot be acted on.
A request needs four items:
- the URL of the page
- the value in question
- the primary source that shows the correct value
- the date of that source
A request never changes inclusion, order, tone or coverage. Pages are not sent to buyers or other named parties for review or approval before publication, and no payment is accepted for any change. Where a party's own pages disagree with each other, both statements stay visible on the page with their dates.
Independence from buyers and referral programs
DataPacts is not a broker and takes no cut of any data sale. As of 2026-11-09 the site carries no referral codes or tracking links. The page on how DataPacts is funded and what it never does holds the full statement.
Authorship and bylines
Pages are published under the byline DataPacts Editorial Team, and no page claims to be written or reviewed by a named expert. The byline note on each page points to the sourcing method, so a reader can see how the figures were gathered and dated.
Related trust pages
The policy works together with the sourcing method, the About page and the contact route.
- How we source, date and check every figure
- About DataPacts, an independent reference
- Contact the editorial team