AI Model Transparency

Version 1.0 Effective: June 3, 2026 Updated when models change

TL;DR — The Short Version

Skylina uses external AI models — we don't train our own. Each model is listed here with what it does, its known limitations, and what training data it uses (per the provider's own disclosures). We don't pretend to know more than we do.

Claude (Anthropic)

Code Generation
Provider
Anthropic PBC
Use in Skylina
Code generation, app scaffolding
Training Data
Anthropic's training corpus (see their documentation)
Last Updated
Per Anthropic's model release notes
CapabilityKnown Limitation
Multi-language code generationMay generate outdated API patterns
Code explanation and reviewCannot execute or test generated code
Bug identificationMay miss edge cases or subtle security issues
Architecture suggestionsRecommendations may not match your infrastructure
User responsibility: You are responsible for reviewing generated code before use in production. Code may contain security vulnerabilities, bugs, or outdated library versions. Test all code before deployment.

For Anthropic's full model card and safety policies: Anthropic Policies

Sapiom Image Generation

Image Generation
Provider
Sapiom (via Polsia)
Use in Skylina
Image and artwork generation
Training Data
Per Sapiom's published model documentation
Watermarking
C2PA-compliant metadata embedded
CapabilityKnown Limitation
Text-to-image generationText rendering in images may be inaccurate
Style varietyMay reflect biases from training data
High-resolution outputQuality depends on prompt specificity
Copyright note: Generated images may inadvertently reproduce elements from training data. You are responsible for ensuring your use of generated images does not infringe third-party rights.

Sapiom Music Generation

Music Generation
Provider
Sapiom (via Polsia)
Use in Skylina
Original music and audio track generation
Training Data
Per Sapiom's published model documentation
Output Format
MP3, WAV

Music outputs are original compositions generated by the model. Generated music may not be suitable for synchronization licensing (film/TV/advertising) without additional clearance.

Sapiom Video Generation

Video Generation
Provider
Sapiom (via Polsia)
Use in Skylina
Short-form video generation from text/image prompts
Training Data
Per Sapiom's published model documentation
Watermarking
C2PA-compliant metadata embedded
Video authenticity: Video content may be visually convincing. Users must clearly label AI-generated video as AI-generated when sharing publicly. C2PA metadata is embedded but may not be visible to viewers.

Model Change Policy

When we switch to a materially different AI model (different provider, significantly different capabilities, or known training data changes), we update this page within 30 days. Minor model updates (bug fixes, efficiency improvements from the same provider) are noted in our Transparency Report.

Training Data Disclosure

Skylina does not train its own AI models. All generation uses external model providers. For each model, training data details are controlled by the respective provider. We link to provider documentation above. For Anthropic's Claude: training data details are published at docs.anthropic.com.

EU AI Act Compliance

Skylina falls under the EU AI Act as a provider of a general-purpose AI system (GPAIS). We comply with GPAIS obligations including: (a) transparency about model capabilities and limitations; (b) technical documentation per Annex VIII; (c) copyright policy for training data; (d) energy consumption disclosure where applicable. Model cards above constitute part of our transparency obligations.

High-risk AI system classification: Skylina does not fall under EU AI Act Annex III (high-risk AI systems) as it does not target fundamental rights applications listed therein.

Contact

Questions about our AI models: ai-transparency@skylina.polsia.app

© 2026 Skylina, built with Polsia.