Public does not mean unrestricted
Free to download and free to do anything with are different things. Five questions worth answering before you build something on a public dataset.
"It is public data" gets used as though it settles the question. It does not. Public means you can get it. What you may then do with it is a separate matter, decided by a licence, and the licences differ.
This is not a legal note and we are not lawyers. It is the set of questions worth asking before you build anything commercial on someone else's dataset.
Five questions
1. What licence is it under, exactly?
Not "it is government data". A named licence, with a version. Many UK public datasets sit under the Open Government Licence, which broadly permits commercial reuse provided you attribute the source. Others do not, including some that are free to download. Free and openly licensed are not the same thing.
2. Can you redistribute it, or only use it?
Some licences let you analyse data internally but not republish the rows. That is the difference between building a report and selling a file, and it is the question most people skip.
3. Does the licence follow the data into your output?
If you combine three datasets, the most restrictive licence tends to govern what you can do with the result. A single restrictive source can constrain an entire product.
4. What attribution is required, and where?
Usually a specific form of words, sometimes in a specific place. It is a low bar and it is a condition, so failing it is a breach even when everything else is fine.
5. Does it contain personal data?
Licensing and data protection are different regimes and you have to clear both. A licence permitting reuse does not authorise you to process personal data however you like. Company officers are individuals, and a name plus a date of birth plus an address is personal data whatever register it came from.
Where people go wrong
The common failure is not malice, it is assumption. Someone downloads a file, sees no paywall, and treats absence of cost as absence of terms. Then a product gets built, customers arrive, and the licence question surfaces at exactly the moment it is expensive to answer.
The second failure is inheriting a restriction without noticing. One enriched field from a restrictively licensed source can quietly change what you are allowed to sell.
What we do about it
Every dataset we sell says where it came from, and we read the licence for a source before building a product on it rather than after. Where a source restricts what a buyer may do, that restriction appears on the product page rather than in a link nobody opens.
Our own terms are on the pricing and licensing page, and the terms of sale set out what you can do with a file once you have bought it.
If you are unsure about a specific source, the answer is to read the licence for that source. It is usually a page long and it is usually clear.
Looking for the data behind this?
Every dataset says where it came from, what is in it and how often it refreshes.