SPT SPT MODELS / LEGAL INFORMATION
AI and transparency
Third-party models, human oversight and identification of synthetic content.
What SPT Models provides
SPT Models is a general-purpose platform for deploying and operating different models, covering text, speech, images, video and other capabilities according to the catalogue and available resources. The platform organises access and execution. It does not automatically make a model suitable for every context or validate it for a regulated profession.
This page provides transparency information and precautions for using the platform. It is not a certification, a general declaration of compliance or a promise that every model and use case meets every requirement of the European AI Act.
Assess roles and the use case
Regulation (EU) 2024/1689 distinguishes roles and obligations that depend on the facts. The roles of Sponge Theory, the model supplier, the integrator and the operator must be assessed for each configuration. Integration, modification or marketing under another name may change that assessment.
Before deployment, document the purpose, affected people, context, decisions influenced and consequences of error. High-risk uses and activities subject to sector-specific rules require a specific assessment. Applicable deadlines and duties must be checked against current legislation; this page announces no general compliance date.
Separate licences and rights
Each model, set of weights, library and resource may have different conditions. Check the licence of the downloaded version, commercial restrictions, attribution duties and rules applying to inputs and outputs. Catalogue inclusion does not grant permission to redistribute a model or sell all its results.
Obtain the permissions needed for voices, faces, documents and other material supplied to a model. Generated output is not guaranteed to be unique, accurate or free of third-party rights. Its legal status depends on the circumstances. Retain relevant references and assess the rights before publication or reuse.
Exclude prohibited uses
Review legal and contractual restrictions before enabling a scenario. AI practices prohibited by applicable rules are not authorised by these pages, a software licence or the technical availability of a feature. Fraudulent activities, abusive exploitation of people and unlawful privacy intrusions must be excluded.
Controlling use requires organisational decisions: access rights, training, project review and incident follow-up. Any filters are a limited supporting measure. We do not claim that a technical mechanism detects or prevents every prohibited use, or that local deployment removes the need to assess lawfulness.
Human oversight and staff competence
Select people who can understand model limitations, verify results and stop problematic processing. Provide human review proportionate to the consequences, particularly before action affecting a person. Test errors, bias and misuse in the intended context; a successful demonstration is not evidence of general safety.
Maintain appropriate documentation of versions, settings, permitted data, known limitations and risk controls. Train operators on these matters and reporting routes. Useful logging must remain compatible with data minimisation and retention policies. Software assistance does not replace these deployment responsibilities. Where a model or workflow changes, reassess the assumptions relevant to the affected use case.
Nora and synthetic media
Nora is a fictional avatar; her image, voice and video are AI-generated. She presents SPT Models and does not represent a customer, a real employee or an authentic testimonial. This information must be visible at first exposure to the media and accompany relevant reuse; a legal page alone is not enough.
For your own synthetic media, assess transparency obligations for the content and its distribution. Article 50 distinguishes, in particular, informing people interacting with AI, machine-detectable marking of synthetic outputs by the providers concerned, and disclosure of certain content by deployers, subject to applicable conditions and exceptions. A visible “AI-generated” label does not replace technical marking where required. SPT Models does not promise universal marking of every output from every model; the technical chain and each participant’s obligations must be assessed for the intended use.
To report an incident, confusion about a media item or a rights infringement, contact contact@sponge-theory.ai with the page, version and a minimal description. Avoid supplying sensitive content unless a suitable exchange method has been agreed. The operator must organise incident handling for its deployment and make any notifications required by applicable law. Reporting an issue to Sponge Theory does not replace those steps.