Mora / How it stays private
How it stays private
A company licenses its records without saying who it is. A lab buys without saying what it trains on. Mora sits between them, and this page says exactly what each one sees.
Two sides, no names
Neither side ever learns who the other is. Not while they talk, not after they agree: each one signs with Mora.
Provider 181C
- company
- status
- Verified company
- listing
- Commercial HVAC service calls
mora, in the middle- Verifies both companies
- Draws the sample at random
- Checks rights, personal data, public copies
- Signs with each side, names stay hidden
Lab 03C9
- company
- status
- Verified company
- offered
- $150k, non-exclusive, 2 years
Who sees what
Every kind of information on Mora, and the only people it reaches.
| What | The other side | Anyone on the web | Mora |
|---|---|---|---|
| Your company name | Never | Never | Yes, to verify you |
| Your anonymous id (Provider 181C, Lab 03C9) | Yes | On public listings | Yes |
| A listing's description and schema | Yes | If the provider made it public | Yes |
| The random sample | Labs, on the listing | Never | Yes |
| The full dataset | One lab, after it signs | Never | Only to check and sample it |
| What a lab asked for or offered | The provider it is for | Never | Yes |
| Passwords to your tools | Never | Never | Never |
People are replaced before anything leaves
The provider's agent writes the schema and marks every personal field. Each person becomes a generated id, the same id everywhere, so a lab can still follow one customer across tickets without ever knowing who it is.
{
"ticket_id": "T-88412",
"customer": "Dana Whitfield",
"phone": "+1 415 555 0134",
"complaint": "no heat on first floor",
"diagnosis": "cracked igniter",
"called_back": true
}{
"ticket_id": "T-88412",
"customer": "p_7f3a91",
"complaint": "no heat on first floor",
"diagnosis": "cracked igniter",
"called_back": true
}| Field | Type | What it holds | Personal data |
|---|---|---|---|
| ticket_id | id | The ticket | None |
| customer | id | Who called | Replaced by a generated id |
| phone | text | Their number | Removed |
| complaint | text | What they reported | None |
| diagnosis | text | What the technician found | None |
| called_back | boolean | Did they call again within 30 days | None |
The example is made up. The schema is the provider's agent's work: nobody fills in a form. Mora then checks the sample for anything personal that was missed.
What happens to a dataset, step by step
You connect a tool
You authorize on the tool's own page. Mora never sees your password, and you can remove the access from the tool at any time.
Mora draws the sample
At random. The provider does not choose the records, so a lab tests the real thing and not the best ten.
Mora checks it
That you may license it. That nobody is named in it. That it is not already free on Hugging Face, Kaggle or the open web.
Each side signs with Mora
The lab licenses from Mora, the provider licenses to Mora. The full dataset goes to that one lab, and no name changes hands.
What Mora does not do
Said plainly, so nobody finds out later.
It does not label or clean
Preparing records, and removing personal data from them, is the provider's work. Mora checks the result.
It does not sell your name
Not to the other side, not to anyone. A listing shows an id; an offer shows an id.
It does not let agents commit you
An agent can search, describe and prepare. Paying, signing and authorizing a tool are done by a person, on a page.
It does not vouch for free data
Public datasets are shown as their publisher describes them. Mora reads the declared license and checked nothing else.
Need private data?
Search, test a sample, make an offer. Nobody learns what you train on.
Search datasets