FussAI uses Lajkto identity and stores data needed to execute, evidence and perform Contributor agreements. Contract data and signatures are not a public catalogue.
We store the exact CZ and EN wording, account identity, entered identity details, required acknowledgements, drawn signature, date/time, IP address, user-agent, SHA-256 hashes of terms and signature, and a final hash after countersignature. For informal bands, each member/rightsholder confirmation is recorded separately.
Records may include material title, type, private source link, duration, copyright/master/performer-right information, publisher or collecting-society information, AI assistance, model lineage, Contribution Units and review audit notes.
After effective termination, the source is removed from the active dataset where reasonably technically and legally feasible and is not to be re-ingested into new training runs. Contract, accounting, legal-claim and model-lineage records may be retained where a valid legal basis exists. Personal-data erasure requests are assessed separately under applicable law.
A named model of a specific recognisable voice uses a separate explicit addendum. A request to end such a voice model is handled more strictly than an ordinary dataset withdrawal and does not waive rights relating to personal data or identity.
Succession notices, rights claims and other Contributor cases are recorded separately. Sensitive probate documents should not be pasted into free text; verification is handled separately.
FussAI uses one credit balance for its creative generative services. The technical storage remains compatible with the existing Lajkto ledger. For the Creator Commitment calculation, FussAI revenue includes only the value of credits actually consumed by FussAI services, not credits used elsewhere in Lajkto.
Prompts, text, images, video, audio, voice references and other files submitted for a specific chat, analysis, edit or generation are processed to create the output, operate securely and perform necessary technical processing. Ordinary Studio use and private storage of a track are not consent to training. The exception is a music output that the user voluntarily publishes to the Public library and, at publication, actively confirms limited use for evaluation, testing, training, fine-tuning and improvement of FussAI. A private prompt or non-public lyrics input is not automatically included. After removal from Public, the output should not be newly included in future training runs where reasonably technically and legally feasible; models already created are not automatically untrained retroactively.
Reference voice recordings, other audio inputs, sound effects, dubbing outputs and text transcripts in Sound AI are kept in active storage for no longer than 365 days from creation or upload of the specific item. Once that period is reached, the oldest items may be automatically and irreversibly removed. Temporary working uploads may be deleted much sooner. Users should download and back up any output they want to keep. Separate accounting, contractual, transaction and security records that do not contain the audio content itself may have a different legally required retention period. Private voice references and audio inputs do not, merely by using Sound AI, grant permission for model training.
Audio inputs, reference recordings and text are processed only to the extent necessary to perform the selected function, secure the Service and deliver the result. Merely using Sound AI does not grant permission to use private inputs for model training.
For some functions, FussAI may use its own or contractually provided compute infrastructure and technical service providers. Data necessary for a specific task is made available only to the extent required to perform the Service, maintain security and deliver the result. Use of such infrastructure is not by itself consent to train models on the user’s private inputs.
admin@lajkto.eu · Contributor case