Licensing

Rights written for model training.

Three tiers. Every dataset page states exactly which apply. The full draft texts are linked below; versioned final texts follow counsel review.

Request pricing

Draft terms

The summaries below describe the intended structure and have not yet been reviewed by counsel. They are not an offer and do not create a licence.

Tier comparison

Research

Commercial AI

Enterprise

Train, fine-tune, evaluate

Yes, non-commercial and academic

Yes, including commercial models and products

Yes

Model and output ownership

Licensee owns; non-commercial deployment

Licensee owns; commercial outputs

Same, plus IP indemnity

Redistribution of raw data

No; cite dataset

No; contractors under NDA

Affiliates and named contractors

Reconstructable data in model

Prohibited

Prohibited; safe harbour for incidental memorisation

Same

Font-generation models

Prohibited

Term and territory

Perpetual, worldwide

Perpetual or subscription

Exclusivity

None

Optional, for custom or carved-out sets

Attribution

Required

Optional

Use restrictions

RAIL-style responsible use

Lawful use; no biometric or real-person inference

Negotiated

Provenance pack

Manifest, Croissant and EU-summary paragraph

Same

Plus audit access and change notices

Warranty and liability

Chain-of-title warranty; cap at fees paid

Chain-of-title warranty; cap at fees

Cap at one to two times fees; IP indemnity

Pricing

Per subset, from 25,000 assets

Per line or full dataset; request pricing

Custom

Research

Train, fine-tune, evaluate: Yes, non-commercial and academic
Model and output ownership: Licensee owns; non-commercial deployment
Redistribution of raw data: No; cite dataset
Reconstructable data in model: Prohibited
Font-generation models: Prohibited
Term and territory: Perpetual, worldwide
Exclusivity: None
Attribution: Required
Use restrictions: RAIL-style responsible use
Provenance pack: Manifest, Croissant and EU-summary paragraph
Warranty and liability: Chain-of-title warranty; cap at fees paid
Pricing: Per subset, from 25,000 assets

Commercial AI

Train, fine-tune, evaluate: Yes, including commercial models and products
Model and output ownership: Licensee owns; commercial outputs
Redistribution of raw data: No; contractors under NDA
Reconstructable data in model: Prohibited; safe harbour for incidental memorisation
Font-generation models: Prohibited
Term and territory: Perpetual, worldwide
Exclusivity: None
Attribution: Optional
Use restrictions: Lawful use; no biometric or real-person inference
Provenance pack: Same
Warranty and liability: Chain-of-title warranty; cap at fees
Pricing: Per line or full dataset; request pricing

Enterprise

Train, fine-tune, evaluate: Yes
Model and output ownership: Same, plus IP indemnity
Redistribution of raw data: Affiliates and named contractors
Reconstructable data in model: Same
Font-generation models: Prohibited
Term and territory: Perpetual or subscription
Exclusivity: Optional, for custom or carved-out sets
Attribution: Optional
Use restrictions: Negotiated
Provenance pack: Plus audit access and change notices
Warranty and liability: Cap at one to two times fees; IP indemnity
Pricing: Custom

Full texts

Version 0.1-draft. Each text carries its draft status in its header and is delivered inside every pack it governs.

What every dataset licence states

Permitted use and prohibited use
Commercial rights and model-training rights
Redistribution and sublicensing
Derivative-model rights and output ownership
Attribution requirements
Exclusivity, where applicable
Term, territory, warranty and liability cap

Fonts and illustration

Typefaces embedded in SVGZO files remain under their own authors' licences, listed in NOTICES/ with every delivery. All tiers prohibit using the datasets to train font-generation models. Illustration and layout are created and processed by The Doodle Desk's team and network of creators.

Content

All figures, names and captions are written placeholders. No real organisation, person or measurement appears in any asset.