License

 

 Last Updated: November 2026

• "Multimodal Asset Set" means a curated package of digital media assets—comprising high-resolution photos, vector graphics, and video clips—bundled together with corresponding structured metadata files in JSON and CSV formats.
• "Raw Assets" refers to the individual photographic images, illustrations, vector graphics, animations, and video files contained within a Multimodal Asset Set.
• "Structured Metadata" means the machine-readable contextual information accompanying Raw Assets, delivered in JSON (JavaScript Object Notation) or CSV (Comma-Separated Values) format, including but not limited to descriptive captions, taxonomy tags, technical specifications, spatial coordinates, bounding box annotations, and temporal segment markers.
• "AI Model" means any artificial intelligence system, deep learning architecture, neural network, foundation model, multimodal model, large vision-language model (VLM), generative model, or algorithm designed to process, analyze, synthesize, or generate data.
• "Training Use" means the ingestion, parsing, tokenization, feature extraction, embedding generation, pre-training, supervised fine-tuning, reinforcement learning from human feedback (RLHF), evaluation, and benchmarking of AI Models using GreenStock Multimodal Asset Sets.
• "Trained Model Artifacts" means the mathematical weights, parameters, latent spaces, embeddings, neural connectivity values, and architectural configurations resulting from Training Use, excluding any raw or un-embedded GreenStock Raw Assets.
• 
  
• Grant of AI Training License
• Subject to Licensee's full compliance with this Agreement and payment of applicable license fees, GreenStock grants Licensee a non-exclusive, non-transferable, worldwide, non-sublicensable license to:
  1. Ingest and Process Assets: Ingest, parse, store, format, tokenize, and process the Multimodal Asset Sets (including Raw Assets and Structured Metadata) within Licensee’s secure computing environments, cloud infrastructure, or private model-training clusters.
  2. Model Pre-Training and Fine-Tuning: Utilize the Raw Assets and Structured Metadata to pre-train, train, fine-tune, align, optimize, and evaluate AI Models, including multimodal vision-language models, image/video generation models, and object recognition systems.
  3. Embedding and Vector Indexing: Extract visual features, generate vector embeddings, and construct index databases derived from the Multimodal Asset Sets for the sole purpose of developing, operating, and improving Licensee’s internal or commercial AI Models.
  4. Internal Evaluation & Benchmarking: Conduct algorithmic performance testing, bias audits, accuracy assessments, and quality assurance benchmarking across the asset sets.
  5. 
     
  6. 
     
  7. Scope of Permitted Deliverables
     This license permits Licensee to commercialize and deploy AI Models trained on GreenStock Multimodal Asset Sets. Licensee is not required to pay residual royalties or attribution fees on downstream AI Model outputs.
  8. 
     
  9. Metadata and Dataset Specifications
  10. JSON and CSV Integration

      GreenStock delivers its multimodal packages paired with structured metadata formatted in standard JSON schemas and CSV spreadsheets. Licensee is granted the right to parse, index, reformat, and integrate these JSON and CSV files directly into proprietary data pipelines, dataloaders, and model training frameworks.
  11. 
      
• Metadata Modifications
  Licensee may supplement, translate, re-annotate, or clean GreenStock Structured Metadata to optimize compatibility with specific AI architectures. Any such enhanced metadata created by Licensee shall be used solely in connection with Licensee's authorized Training Use.
• 
  
• Proprietary Rights and Model Ownership
• 
  
• Ownership of GreenStock Assets
  GreenStock and its contributing artists retain all rights, title, and interest—including all copyright, trademark, and intellectual property rights—in and to the underlying Raw Assets, original Structured Metadata, and Multimodal Asset Sets. Nothing in this Agreement transfers copyright ownership of the raw photos, videos, vectors, or metadata schemas to Licensee.
•  Ownership of Trained AI Models

  As between GreenStock and Licensee, Licensee retains sole ownership of all Trained Model Artifacts (such as neural network weights, parameters, and model checkpoints) generated through authorized Training Use. GreenStock claims no proprietary interest or copyright ownership in Licensee's independent AI Models, provided they do not reproduce raw GreenStock assets
• 
  
• License Restrictions and Prohibited Uses

  Licensee explicitly agrees to abide by the following operational and commercial restrictions:

  • No Public Stock Resale or Redistribution: Licensee shall not redistribute, sub-license, host, publish, or resell Raw Assets, Structured Metadata, or Multimodal Asset Sets as a standalone stock photo, video, or vector repository, nor make them available to third parties as a digital asset marketplace.
  • No Raw Dataset Exposure: Licensee shall not publish, expose, or publicly distribute the raw JSON/CSV metadata files or un-embedded media assets in a manner that allows third parties to extract, download, or reconstruct GreenStock's original asset packages.
  • No Exact Memorization or Re-Creation: Licensee shall implement technical safeguards (such as deduplication, regularization, and output filtering) to prevent generative AI Models from memorizing and outputting verbatim, high-fidelity replicas of GreenStock Raw Assets as standalone media files.
  • No Unlawful or Harmful Training: Licensee shall not utilize the assets to train models intended for unlawful, defamatory, pornographic, libelous, or illegal purposes, nor for the unauthorized synthesis of fraudulent personal identification documents.
  • No Indicator of Origin: Licensee shall not use GreenStock’s name, trademarks, or logos as an indication of origin for Licensee’s AI products without prior written authorization.
  • 
    
  • Editorial Use Only Content
    Certain Multimodal Asset Sets may contain assets designated as "Editorial Use Only."
    1. Scope of Editorial Assets: Editorial assets include depictions of real-world public figures, unreleased property, identifiable landmarks, or trademarked logos where commercial model releases have not been obtained.
    2. Training Limitations: Licensee may utilize Editorial Use Only assets for model pre-training, computer vision research, object detection, and academic analysis.
    3. Downstream Commercial Output Disclaimer: Licensee assumes sole responsibility for ensuring that its commercial deployment of trained models does not generate outputs that violate third-party publicity, privacy, or trademark rights associated with Editorial content.
    4. 
       
    5. 
       

 Account Management, Security, and Monitoring