An independent standard, published in full.
Agent Certified is a methodology and assessment standard for AI agents in production, published by Future Proof Intelligence. This page sets out where it came from, who stands behind it, and the instruments it is built on.
Two questions, one answer.
The framework was first published in April 2026, when two pressures met on European organisations running autonomous agents in production.
The first was the staged application of the EU AI Act, with its duties on governance, human oversight and post market monitoring. The second was the question insurers face when asked to cover professional indemnity and technology errors arising from decisions an agent made.
Both were asking the same thing in different language. A regulator asks whether the operator can evidence the controls around the agent. An insurer asks whether the risk posture is consistent enough to price. Existing AI standards and management system frameworks did not give either audience a single readable signal.
Agent Certified was designed to give one. It takes its structure from technical standards bodies and its weighting discipline from rating methodologies. The seven dimensions are deliberately few, so the signal stays legible, and deliberately weighted, so it reflects operational reality instead of checklist completeness.
Who publishes it, and what it is.
Future Proof Intelligence publishes and maintains Agent Certified as an independent methodology, and states plainly what that does and does not make it.
What it is
A research and assessment standard producing structured evaluations that operators, boards and counterparties may choose to rely on as part of wider due diligence.
What it is not
A government accredited certification body, a notified body under the EU AI Act, a regulator or a licensed insurer. A result is no substitute for conformity assessment under Article 43.
How it stays independent
The methodology team works independently of any insurer, platform vendor or model supplier. No assessed operator holds an interest in the publisher. Conflicts are declared on the evidence file.
The methodology is published openly. Anyone may read the full rubric, the weights and the scoring descriptions without cost or registration. Future Proof Intelligence also intends to offer assessments under it; that interest is disclosed in the editorial standards.
Anchored to the primary instruments.
The framework does not replace primary regulatory instruments. It is built on them, and these are cited as foundational references in the methodology and in every evidence file.
| Instrument | Issued by | Role in the framework |
|---|---|---|
| EU AI Act, Regulation (EU) 2024/1689 | European Parliament and Council | The primary legal anchor. Articles 9, 10, 14, 15 and 26 map directly to dimension scoring. |
| ISO/IEC 42001:2023 | International Organization for Standardization | Management system structure. Operators aligned to ISO/IEC 42001 will find their controls translate readily. |
| NIST AI RMF 1.0 | US National Institute of Standards and Technology | The risk function model, and the source vocabulary for Trust and Safety and Context Integrity. |
| Opinion on AI governance and risk management, 6 August 2025 | European Insurance and Occupational Pensions Authority | Sector alignment, so the result reads clearly to European insurers looking at agent exposure. |
One of six, read together.
Agent Certified is one of the Future Proof standards: six publications, each taking one part of the same question about trust in the AI era.
| Publication | What it covers |
|---|---|
| Agent Certified | The standard for AI agents in production. This site. |
| Future Proof Certified | The standard for companies built for the AI era. |
| Agent Liability Europe | Who answers when an agent acts, under EU law. |
| Agent Liability Global | The same question, jurisdiction by jurisdiction. |
| Agent Insured | The market for insuring AI agents, followed closely. |
| Insure Your Agent | The operator's guide to getting covered. |
Press, research and enquiries.
Journalists, researchers and standards bodies who want to discuss the methodology or its revision history are welcome. Methodology questions, conflict declarations and appeals arrive the same way. Every request is acknowledged within five working days.
Write to us through the request page, and say what it concerns
Read the method before anything else.
The seven dimensions, their weights and the scoring rubric are published in full.