Who takes part, and on what terms.

An independent methodology and assessment framework for autonomous AI agents in European organisations, aligned with Regulation (EU) 2024/1689, ISO/IEC 42001:2023 and the NIST AI Risk Management Framework.

An organisation that requests an assessment and meets the criteria below receives a certification level. The methodology is published in full at Methodology v2.0, so any reader can inspect the basis of any outcome. That openness is what lets a third party use it for due diligence.

All editorial coverage on this site, including articles, analysis and framework comparisons, is produced on an editorial basis. No carrier, vendor, consultancy or third party pays for placement, ranking or favourable framing.

Assessment work is kept apart from editorial decisions. The commitment is set out in full in the editorial standards.

Any operator that has deployed at least one autonomous AI agent in production, meaning the agent performs tasks with real consequences for real users or on real data. Operators still before deployment may ask for a preliminary readiness review, which is distinct from a certification assessment.

Evidence. An assessment needs evidence across the seven weighted dimensions: governance documents, technical documentation, deployment architecture, incident logs and records of human oversight. An organisation unwilling to provide evidence at the required level of detail cannot score above Pre Assessment.

Geography. The methodology is written for agents deployed within or into the European Union. It can be applied elsewhere, and the result is framed against the European regulatory context; any certification notes where the deployment reaches beyond the EU and where the method was adapted.

A standard or methodology qualifies for our analysis and comparisons if it is published by a recognised standards body (ISO, NIST, CEN, ETSI, BSI or equivalent), an accredited certification body or a supervisory authority, and if it addresses the safety, transparency, governance or accountability of AI systems in a substantive, documented way.

To propose one, send a link to the primary document through the request page with "Editorial inclusion" in the message. Inclusion is editorial and free of charge, and a submission does not guarantee it.

We may introduce a partner programme in 2027 for verified institutions such as carriers, brokers, consultancies and implementation partners. It has not launched, and it has no terms or prices.

If it is introduced, every paid placement will be labelled as commercial. Partnership will not affect assessment outcomes, scores, methodology decisions or editorial content: a partner receives no preference in any assessment, and an institution that stays outside is not excluded from coverage.

Institutions that would like to hear when consultation on the programme opens can say so through the request page. We do not commit to a launch date.

An independent evaluation of one AI agent deployment against published criteria. It is not a government accreditation, a regulatory approval, a legal opinion or an underwriting decision, and Agent Certified is not a conformity assessment body under the AI Act.

Anyone using a result for underwriting, investment or procurement should satisfy themselves that the methodology fits their own requirements.

Future Proof Intelligence publishes this site and intends to operate the assessment service. The dual role is disclosed here and in the editorial standards, and the methodology is published in full so that any outcome can be checked independently.