AEP

Protocol Specification

An open, verifiable and extensible AI evidence protocol

Overview

AEP (Attestation & Evidence Exchange Protocol) is an open protocol that provides trusted digital proof for AI-generated content.

The core problem it solves: in an era of mass AI-generated content, how to prove who generated a piece of content, when it was generated, and whether it remains complete and unaltered.

Through cryptography, time proof and verifiable data structures, AEP builds a complete trusted chain from generation to verification for AI content, independently verifiable by anyone.

Design Principles

Openness

An open protocol; anyone can build their own implementation

Security

Built on proven cryptography; every claim is verifiable

Extensibility

Modular design supporting algorithm and data structure extensions

Interoperability

Compatible with existing standards and ecosystems

Core Concepts

Evidence

A verifiable claim, the basic unit of the protocol

Artifact

The AI-generated content artifact

Event

A record of an operation during generation

Identity

The identifier of the generator

Evidence Graph

A graph of events and artifacts fully recording the generation process

Evidence Levels

From L0 to L7, evidence strength increases progressively

L0

No evidence

L1

Claim: content only

L2

Hash record: content fingerprint notarized

L3

Signed record: signature + hash, traceable source

L4

Timestamp: adds time proof, replay protection

L5

Event chain: continuous event records, auditable process

L6

Complete graph: closed evidence graph

L7

Judicial grade: full evidence chain for legal scenarios

Relationship with Other Standards

C2PA

C2PA focuses on content provenance (Content Credentials); AEP focuses on verifiable evidence of the generation process. They complement each other.

W3C PROV

PROV is a general provenance standard; AEP's evidence graph is compatible with its model, focusing on AI generation scenarios.

MCP

MCP is a protocol for AI-tool interaction; AEP provides MCP integration for native AI agent support.