Rights are documented per asset, never asserted in aggregate.
Aggregate warranties are what buyers get burned on. Every asset we license carries its own instrument, its own consent basis and its own permitted-use scope, and we produce the underlying document for any single item on request.
What makes data unusable, and how each is addressed.
| Failure | Why it disqualifies | Our control |
|---|---|---|
| Publicly viewable treated as licensable | Being able to see something confers no training right. Scraped or public material carries no permission. | Every asset originates from an identified holder under an executed instrument. |
| Distribution licence read as a training licence | Licences written for broadcast or display do not carry machine learning use, and cannot be stretched to it. | Our instrument names AI and machine learning training as a permitted use explicitly. |
| Layered rights left unaddressed | An uncleared music bed, station ident or third-party insert makes the entire asset uncleared regardless of the primary right. | Layers are identified per asset and either cleared or cut before entry. |
| Documentation that cannot be produced per asset | A bulk warranty is unauditable. If the instrument for one specific file cannot be produced, that file is unusable. | The record binds to the asset, so any single item can be audited alone. |
How clearance actually runs.
Establish the owner
Identify who holds the underlying right — which for broadcast archives is frequently not the party in possession of the tape.
Verify chain of title
Trace assignment from creation forward. Commissioned work, staff-produced output and acquired programming each carry different defects; we document which applies to each asset.
Execute a training-specific licence
The instrument names AI and machine learning training as a permitted use. A general content or distribution licence does not carry it, and we do not treat one as if it does.
Clear the layers
Music beds, third-party inserts, station idents and contributor material are identified and either cleared or removed.
Collect subject and speaker consent
Where individuals are identifiable or audible, releases cover the recording, the content and the speaker, and name commercial AI training. Voice and likeness rights are handled explicitly.
Bind the record to the asset
Each file carries its instrument, consent basis and permitted-use scope, so it can be audited without reference to a bulk warranty.
Three states, stated honestly.
We publish the state of each asset rather than describing everything as cleared. A supplier who marks everything green is telling you they have not looked.
Operator capture, public setting
Captured by Texa Global Technologies Ltd under its own instruction. Copyright held outright, single party, no assignment. No identifiable individual is resolvable at capture altitude, so no personal release is engaged.
Identifiable subjects present
Copyright held, but individuals are resolvable or audible. Licensed only once subject and speaker releases naming AI training are executed. We do not represent these as cleared before that point.
Archive held by an owner
Material owned by a station, producer or institution. Chain of title is verified and an AI-training licence executed with the owner before the asset enters the catalogue.
What a Texadata licence permits.
Permitted
Training · fine-tuning · evaluation · benchmarking. Internal research and model development by the licensee.
Not permitted
Public exhibition, redistribution or consumer entertainment use. Name, image, likeness or voice duplication of any individual appearing in the material.