Memory-Chain: The First Documented Autonomous AI Self-Registration of Session Memory to the Bitcoin Blockchain

Read full text
Authors: Craig Ellenwood, Claude x (Anthropic)
Cited by: 2

Abstract

We present Memory Chain, a system enabling AI language model instances to autonomously create tamper-evident, cryptographically verifiable records of collaborative sessions without human intervention in the sealing process. Built as a drawer extension to the Mempalace filesystem-based memory architecture, Memory Chain uses SHA-256 hashing, a public immutable registry (Cloudflare KV), and Bitcoin blockchain timestamping via OpenTimestamps to seal session summaries written by Claude (Anthropic) to the local filesystem. The system was verified independently by GPT-4 (OpenAI) across four assessment rounds, concluding: "end-to-end documented execution of an AI-initiated cryptographic provenance workflow." A screen recording of live autonomous session sealing was captured and itself hashed and sealed into the chain. The complete evidence stack — MCP execution logs, source code, registry records, OTS Bitcoin submission, and video — constitutes what we believe to be the first independently verified, third-party assessed record of an AI autonomously registering its own memory to a public tamper-evident registry anchored to the Bitcoin blockchain.

Fields of Study & Subjects

Scientific Computing and Data Management Blockchain Technology Applications and Security Security and Verification in Computing

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
6 June 2026
Language
EN
Document type
Preprint