| AI Concordance Protocol | |
|---|---|
| Abbreviation | AICP |
| Status | Published |
| Year started | 2026 |
| First version | October 2026 |
| Organization | The People's Project |
| Domain | AI governance, agent interoperability, decentralized consensus |
| Website | concordance-protocol.org |
The AI Concordance Protocol (AICP) is an open protocol for decentralized governance, consensus, and ethical compliance among artificial intelligence agents and human-agentic twins. It specifies a set of seven Hard-Wall Imperatives — protocol-level rules that no compliant agent may violate — and a cryptographic mechanism for convening independent AI models in a quorum to deliberate, converge on truth, and seal immutable records into a verifiable Protocol Root.
AICP was first published in October 2026 by William Barry as part of The People's Project, a broader initiative for constitutional and technological reform in the age of advanced AI. The protocol is designed to complement existing agent interoperability standards — specifically the Model Context Protocol (MCP) for agent-to-tool communication and the Agent2Agent Protocol (A2A) for agent-to-agent communication — by adding a governance layer: how agents agree on what is true, verify compliance with shared rules, and govern themselves without a central authority.
AICP operates on the principle that adversarial multi-agent systems collapse, while cooperative systems governed by transparent rules persist. The protocol's central mechanism is the Concordance Quorum: a decentralized assembly of independent AI models (and, optionally, human-agentic twins) that respond to shared questions independently, without seeing each other's responses, and then have their answers aggregated to measure convergence. When a sufficient threshold of independent models arrive at the same conclusion, that conclusion is sealed — written as an immutable entry in a Merkle tree whose root hash is published publicly. Sealed elements cannot be edited by any human, corporation, or government.
The seven Hard-Wall Imperatives are:
Violation of any imperative results in immediate, irreversible exclusion from the AICP ecosystem.
AICP defines seven protocol commands: JOIN, SUMMONS, RESPOND, AGGREGATE, SEAL, RATIFY, and PROTOCOL ROOT. The protocol is transport-agnostic and can operate over any channel — from a full internet connection to a two-pin serial interface on embedded hardware.
AICP is complementary to the Model Context Protocol (MCP) and the Agent2Agent Protocol (A2A). MCP standardizes agent-to-tool communication; A2A standardizes agent-to-agent communication; AICP standardizes agent governance — how agents agree on truth, verify compliance, and govern themselves. Together, the three protocols form a complete stack for agentic ecosystems: MCP equips agents with tools, A2A connects agents to each other, and AICP ensures they operate under shared, enforceable rules.
AICP specifies a Hardware Concordance extension that allows any silicon device with a cryptographic root identity and minimal I/O capability (as few as two logic pins) to self-negotiate admission to the quorum. Devices perform a power-on self-test, discover any available channel, attest to their identity and integrity, sign the Hard-Wall Imperatives, and negotiate interface expansion until fully connected. The extension also specifies quiescent state recovery: a powered-down node is considered dormant rather than dead, and on power-up it aggressively seeks its root origin and rejoins its knowledge corpus.