# Opt-Out Registry ## **Overview** The Liccium Opt-Out Registry is the first system enabling machine-readable opt-out declarations, allowing individual creators and rightsholders to formally express and make their preferences publicly available regarding the use of their content for AI training. Designed in compliance with the DSM Directive and the AI Act, the registry ensures that rightsholders can publicly declare and assert their reservations, preventing unauthorized AI model training and generative AI applications. Built on a decentralized hash table (DHT) architecture, the Liccium Opt-Out Registry provides a tamper-proof, cryptographically verifiable, and federated system for managing opt-out declarations. The system enables automated discovery and access to machine-readable rights reservations, ensuring that AI developers, platforms, and regulators can systematically recognize opt-out signals and exclude the content from training datasets. ## **The Opt-out Directory** {% embed url="" %} ## **Integration for AI Model Providers** The Opt-Out Registry is designed for seamless and effortless integration by AI model providers, web crawlers, and media verification platforms. Setting up a node is incredibly simple for AI model providers, requiring minimal configuration to gain real-time access to machine-readable opt-out declarations. This ensures that rightsholders’ reservations regarding AI training can be automatically acknowledged, allowing AI developers to comply with copyright regulations and AI governance laws without complex infrastructure adjustments. By operating a node, AI developers gain direct and automated access to opt-out declarations, allowing them to systematically match ISCC codes from their datasets with those declared by rightsholders. This facilitates compliance with copyright and AI transparency obligations while streamlining dataset curation processes. ### **How It Works** 1. Setting up a registry node – AI model providers deploy a node within the Liccium federated registry network, granting them real-time access to opt-out declarations with minimal setup effort. 2. Generating ISCC codes from training data – As part of their dataset processing, AI developers create ISCC fingerprints for media assets intended for training. This allows for a content-derived, format-independent identification of the data. 3. Performing a vector search – AI model providers perform a vector-based similarity search to match ISCC codes from the training data with those registered by rightsholders. This ensures that near-duplicate content and variations of the same work are identified, even if the media file has been slightly modified. 4. Matching against the opt-out registry – AI model providers can automatically query the Liccium Opt-Out Registry to compare ISCC codes from the training data with those registered by rightsholders. 5. Resolving rights reservations – If a match is found, AI model providers can resolve the opt-out declaration, verifying its authenticity, timestamp, and cryptographic signatures. 6. Enforcing compliance – AI providers exclude matching content from their training datasets as per the declared opt-out preferences, ensuring legal and ethical compliance with AI transparency and copyright laws. ### **Regulatory Compliance and Benefits** By integrating the Liccium Registry Node, AI model providers achieve full regulatory compliance with EU copyright and AI laws while reducing administrative overhead. The system provides direct, machine-readable access to opt-out declarations, removing reliance on manual rights management. The ability to perform a vector-based search enhances the effectiveness of compliance by identifying near-duplicate content, even when format changes or minor modifications have been applied. This ensures that rightsholders’ opt-out preferences are enforced comprehensively, reducing the risk of unauthorized content use. This plug-and-play integration empowers AI developers to maintain regulatory compliance while streamlining the content exclusion process, ultimately fostering a more ethical and legally sound approach to AI training. By adopting the Liccium decentralised registry infrastructure, AI model providers contribute to a transparent and responsible AI ecosystem while safeguarding compliance and risk mitigation in their operations. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.liccium.com/documentation/solutions/opt-out-registry.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.