AITF is building the open cryptographic identity standard for AI agents, GPUs, and AI models — the way DNS works for websites, AITF works for agents.
AI agents are taking autonomous actions — moving money, making decisions, executing code — with no cryptographic proof of who they are or what they are authorised to do. AITF exists to fix this, permanently, as an open standard.
Every AI agent and GPU deserves a post-quantum signed cryptographic passport. Not a username. Not an API key. A verifiable, unforgeable identity bound to the agent itself.
Every action taken by an agent must be signed, timestamped, and logged immutably. EU AI Act Article 12 is law today. Accountability infrastructure must be open and universal.
Agent identity cannot be owned by a single company. AITF develops and maintains an open standard — free to implement, free to audit, governed by the community.
The AITF standard uses CRYSTALS-Dilithium3 and Kyber768 — NIST-selected post-quantum algorithms. Agent identities issued today will remain valid in the quantum era.
The standard maps directly to EU AI Act Article 12, NIST AI RMF, SOC2, and ISO 27001. Compliance is not an afterthought — it is baked into the protocol layer.
AITF is governed by its members — security companies, AI infrastructure providers, universities, and enterprises. No single entity controls the standard.
The AITF standard defines a cryptographic identity protocol for AI agents, GPUs, and AI models.
A post-quantum signed X.509 certificate bound to each AI agent. Contains identity, capability scope, issuing organisation, and validity period.
X.509 certificate issuance for GPU and hardware devices. TPM 2.0 ready. Hardware-bound signing with SOC2-ready audit logs.
Distributed certificate authority using MPC — n-of-m quorum required to sign. No single private key. No single point of compromise.
Every agent action signed with its passport and written to an immutable audit log. Directly maps to EU AI Act Article 12 requirements.
OPA-compatible policy language for defining agent capability scopes. Runtime enforcement before every action.
ComputeID is the production reference implementation of the AITF standard — built by the foundation's founding team. Live API, 3 PyPI packages, 13 MCP tools, and a public passport issuance portal. Try it at passport.aicomputeid.com.
Everything is open source, free to use, and MIT licensed. Install and integrate in minutes.
Python SDK for issuing, verifying, and revoking agent passports.
pip install computeid-sdk
Command-line tool for identity management and audit log access.
pip install computeid-cli
13 MCP tools natively usable by Claude and other agent frameworks.
pip install computeid-mcp
Live REST API for all core operations. Full OpenAPI spec available.
api.aicomputeid.com
Issue and verify agent passports without any setup or login.
passport.aicomputeid.com
All source code, specifications, and documentation publicly available.
trustedaicompute-ops
AITF follows a phased governance model — from foundation-led open source to a fully community-governed, decentralised standard body.
Open source release. Reference implementation live. Community building. First enterprise adopters. EU AI Act compliance framework.
Draft RFC for "Agent Identity and Capability Attestation". Academic partnership with KU Leuven COSIC. Working group formation.
Join or establish IEEE working group for agent identity standards. Regulatory alignment across EU, US, and APAC jurisdictions.
Distributed node operators. Community-governed CA. Every major agent framework uses AITF by default. The standard becomes invisible infrastructure.
Whether you are an enterprise deploying AI agents, a security researcher, a framework developer, or a regulator — AITF needs your participation to become the universal standard.