ERC-7007: Verifiable AI-Generated Content Token
Verifiable AI-Generated Content Token
ERC-7007 is an extension of the ERC-721 non-fungible token (NFT) standard, specifically designed to accommodate verifiable AI-generated content on the Ethereum blockchain.
It can be paired with the Initial Model Offering (IMO) in order to create inference assets that benefit model token holders.
What is ERC-7007?
ERC-7007 is a token standard that enables the creation and management of NFTs representing AI-generated content. It introduces a set of interfaces and events that allow for:
Verification of AI-Generated Content: Ensures that the content associated with an NFT is generated from a specific AI model using a given input (prompt).
Integration with zkML and opML: Supports Zero-Knowledge Machine Learning (zkML) and Optimistic Machine Learning (opML) to verify the correctness of outputs.
Unique Token Identification: Assigns token IDs based on the prompt, ensuring that each NFT is uniquely associated with its input and output.
Why is ERC-7007 Important?
As AI-generated content becomes more prevalent, there is a need for a standardized way to represent and verify this content on the blockchain. ERC-7007 addresses several challenges:
Verification of Authenticity: Allows users to verify that the AI-generated content is produced by a specific model with a specific input.
Support for AI Model Authors and Content Creators: Provides a way for creators to monetize their AI models and content without risking open-source devaluation.
Enabling Revenue-Sharing Mechanisms: Facilitates fair compensation for AI model authors and content creators through secure and verifiable NFTs.
How ERC-7007 Works
Core Components
AI Model: A pre-trained machine learning model that generates output based on a given input (prompt).
zkML / opML:
zkML (Zero-Knowledge Machine Learning): Uses zero-knowledge proofs to verify that the content was generated correctly without revealing the model's details.
opML (Optimistic Machine Learning): Uses fraud-proofs where correctness is ensured through a challenge period and and onchain dispute resolution.
AIGC-NFT smart contract: An onchain contract compliant with ERC-7007, with full ERC-721 functionalities
Verifier smart contract: implements a verify function. When given an inference task and its corresponding ZK proof or opML finalization, it returns the verification result as a boolean.
Workflow Overview
zkML Scenario
Model Publication: The AI model and its zero-knowledge proof verifier are published on Ethereum.
Inference Task: A user submits an input (prompt) and initiates an inference task.
Proof Generation: Nodes maintaining the model generate the output and a zero-knowledge proof of the inference.
Verification: The proof is verified on-chain using the verifier contract.
Minting NFT: The user owns the NFT representing the AI-generated content, secured by the proof.
opML Scenario
Model Publication: The AI model is published on Ethereum using the ORA network.
Inference Task: A user submits an input (prompt) and initiates an inference task.
Result Submission: Nodes perform the inference and submit the output.
Challenge Period: Other nodes can challenge the result within a predefined period.
Finalization: After the challenge period, if unchallenged or successfully defended, the user can verify ownership and update the AIGC data as needed.
Metadata Schema
ERC-7007 introduces a standardized JSON schema for NFT metadata, including:
name: Identifies the asset.
description: Describes the asset.
image: URI pointing to an image representing the asset.
prompt: The input used to generate the AI content.
aigc_type: Type of AI-generated content (image, video, audio, etc.).
aigc_data: URI pointing to the AI-generated content.
proof_type: Indicates whether it's a validity proof (zkML) or fraud proof (opML).
Examples
Use Case: AI Art NFTs
An artist uses an AI model to generate artwork based on user-specified prompts:
Model Deployment: The artist deploys the AI model on Ethereum using the ORA Network.
Content Generation: Users submit prompts during the minting process to generate unique artworks.
Verification: Each piece is verified to confirm it was generated by the specific AI model using the provided prompt.
NFT Minting: The verified art is minted as an ERC-7007 NFT, ensuring authenticity and uniqueness. Proceeds from the mint flow back to the artist.
Use Case: AI-Generated Music
A musician tokenizes an AI model using an Initial Model Offering (IMO) that generates music tracks:
Model Publication: The AI music model is tokenized using the IMO process.
User Interaction: Fans submit prompts or themes to generate personalized music tracks.
Verification and Minting: Tracks are verified and minted as NFTs using ERC-7007 and ORA’s AI Oracle.
Ownership and Royalties: The musician can implement royalty mechanisms for each NFT sale. Additionally, the use of the IMO allows token holders to receive a corresponding portion of the mint fees.
Additional Resources
ERC-7007 Proposal: Ethereum Improvement Proposal
Related Concepts:
Last updated