FAQ
Frequently asked questions
Short answers about verifiable AI, AIEF execution artifacts, and how they relate to audits, agent identity, and conformance.
- What is AIEF?
- AIEF (AI Execution Framework) is an open specification for verifiable AI execution evidence — tamper-evident records of AI decisions that can be verified deterministically, independently of the system that produced them.
- What is verifiable AI?
- Verifiable AI means a third party can confirm what an AI system actually decided and on what inputs, using a structured execution artifact whose integrity is checked deterministically. AIEF defines the minimum contents and verification rules for such an artifact.
- How is AIEF different from AI observability or logging?
- Logs and traces are operational telemetry produced and held by the same system being audited. AIEF artifacts are designed as evidence: their protected fields are tamper-evident, the verification rules are public, and any compliant verifier can check them.
- Does AIEF require deterministic AI models?
- No. AIEF treats the recorded output of the model as authoritative for the artifact. It verifies that the recorded decision has not been modified since issuance — not that the model itself is deterministic.
- How can merchants verify AI agent identity during checkout?
- Merchants accepting payments or actions from autonomous AI agents can require a signed AIEF execution artifact for each agent action. The artifact records who issued the decision, the inputs it acted on, and a tamper-evident hash chain. See the agent identity guide for the end-to-end pattern.
- What are AIEF conformance levels?
- AIEF defines conformance levels (L1, L2, L3) that describe progressively stronger guarantees about artifact integrity and stability. Level 2 and Level 3 are appropriate for production evidence used in audits or disputes.
- Where is the canonical AIEF specification?
- The canonical source is the AIEF repository on GitHub (github.com/artnames/AIEF). This site renders each released version at a stable URL under /spec/<version>.
- Is AIEF a governance, safety, or fairness framework?
- No. AIEF is an evidence framework. It does not govern model behaviour, fairness, alignment, or correctness — it specifies how a decision record can be produced and verified.
- What is agentic AI identity?
- Agentic AI identity is the ability to establish which autonomous agent performed an action, under whose authority, and on what inputs. AIEF expresses this as a signed execution artifact bound to the issuing system, so identity claims can be checked by a relying party rather than trusted on assertion.
- How is AI agent identity different from human identity?
- Human identity authenticates a person once and infers intent from session context. An AI agent acts continuously and at machine speed, so each action needs its own evidence: what was decided, on which inputs, and under which delegated authority. AIEF binds that evidence to the individual execution rather than to a login session.
- How do agentic AI payments prevent fraud?
- Fraud controls for agentic payments depend on being able to prove that a transaction originated from an authorised agent acting within its mandate. A signed AIEF execution artifact accompanying the transaction lets the merchant or processor verify origin, inputs, and integrity deterministically, and retain that evidence for chargebacks or disputes.
- How do businesses integrate agentic AI payments safely?
- Require an AIEF execution artifact for each agent-initiated action, verify it at the point of acceptance using the published verification rules, and store the artifact alongside the transaction record. Verification is offline and does not depend on the agent's originating platform remaining available.
- How do you audit an AI agent's decisions?
- Collect the AIEF artifacts issued for each decision and re-run verification independently. Because protected fields are tamper-evident and the verification rules are public, an auditor can confirm the recorded decisions are unmodified without access to the producing system.