AI Transparency Notice
Effective: August 11, 2026
Version: 1.2
EVELI is an AI-assisted creative-production service. Images and other content generated through EVELI are synthetic or materially AI-assisted unless the interface clearly says otherwise.
1. What EVELI does
EVELI uses generative AI models and structured user selections to create faces, bodies, anchor packs and other production assets. The system may use third-party models and EVELI-developed orchestration, taxonomies, workflows and quality controls.
AI generation is probabilistic. Similar requests may produce different results, and different users may receive similar results. Outputs can contain errors, artifacts, unintended similarities or implausible details.
2. Human review
EVELI does not promise that Outputs are factually accurate, rights-cleared, unique or appropriate for a specific campaign. Users must review Outputs before use, especially for advertising claims, real-person resemblance, product accuracy, cultural sensitivity, trademarks and regulated contexts.
3. Machine-readable marking
EVELI's configured live image route is fal.ai's \fal-ai/nano-banana-pro\ and \fal-ai/nano-banana-pro/edit\, which fal.ai identifies as Google's Gemini 3 Pro Image and states includes SynthID marking on generated Outputs. EVELI's durable-storage path copies the provider's output bytes into private storage without image re-encoding. Before EU access is enabled, EVELI will test a final customer download with an available detector and retain the sample, model route, date and result. If the marking does not survive the actual export path, EU generation will remain disabled until an effective machine-readable marking is added.
Users may not intentionally remove, disable, corrupt or falsify required provenance. Format conversion, screenshots and third-party platforms may affect metadata; users remain responsible for any visible disclosure required in the final context.
4. Visible disclosure and deepfakes
The product identifies the workflow as AI generation and this Notice applies to every EVELI Output. Before EU access is enabled, the download surface will also carry a concise “AI-generated” label. If an Output appreciably resembles a real person, object, place, entity or event and could falsely appear authentic, the user must make a clear disclosure wherever required by law or necessary to avoid material deception.
EVELI does not authorize non-consensual deepfakes, deceptive endorsements or false evidence. Creative, satirical or fictional contexts do not remove applicable disclosure, consent or rights obligations.
5. No biometric identification or high-impact decisions
EVELI does not offer face identification, identity matching, emotion recognition or biometric categorization. It is not designed for decisions about employment, credit, housing, insurance, education, medical care, law enforcement or access to essential services.
6. Reporting
Report missing provenance, deceptive use or AI-safety concerns to legal@eveli.ai with the Output or asset identifier and relevant evidence.
7. Role, coverage and regional disclosures
EVELI conservatively operates as the provider of the branded EVELI AI service and as a deployer/user of the underlying fal.ai/Google model route. It therefore applies the provider-side marking and user-facing disclosure controls described above for EU distribution. EU AI Act Article 50 obligations apply from August 2, 2026; EVELI does not rely on the limited transition for systems placed on the market before that date.
The separate AI Training Data Transparency page contains EVELI's repository-verified disclosure for the August 11, 2026 release. EVELI will review and update it before a substantial modification is released to California. EVELI also monitors whether its monthly traffic reaches any California threshold that triggers additional generative-AI detection or manifest-disclosure duties.