GRDprocess · Exploratory study · 2026

How Do We Really Evaluate AI?

An Exploratory Survey on AI Evaluation Practices — 2026

This study examines how organizations assess AI systems in practice: what they evaluate, how requirements are verified, how responsibilities are assigned, and which risks or evaluation gaps remain insufficiently addressed. We welcome perspectives from organizations that conduct, commission, or undergo AI evaluations.

Participate in the survey

The questionnaire covers evaluation practices, behavioral robustness, governance, due diligence, and emerging challenges. Some questions invite optional written comments.

Participation does not require contact information. Communication preferences at the end of the questionnaire are optional and independent. Please do not include confidential or personally identifiable information in free-text responses.

How Do We Really Evaluate AI?

Organization & Context

This section helps us understand your organization's activities and involvement in AI evaluation. An organization may both conduct evaluations and be evaluated by others.


AI Evaluation Practices

This section explores when AI evaluations take place, what they cover, and how they are conducted. Please consider evaluations your organization performs, commissions, or undergoes.


Third-Party AI & Behavioral Robustness

This section examines risks associated with third-party AI models, autonomous agents, and unexpected AI behavior, including situations where systems operate within their instructions but produce unintended outcomes.


Governance, Accountability & Responsibility

This section explores how organizations assign responsibilities for AI systems, make decisions, and address accountability, including when third-party AI technologies are involved.


Due Diligence & Verification

This section explores the evidence, standards, and verification practices involved in AI-related investment decisions, vendor assessments, and procurement processes. Questions apply whether your organization requests, provides, or reviews evaluation evidence.


Challenges, Gaps & Emerging Needs

This final section explores the practical challenges organizations encounter in AI evaluation, the circumstances in which independent assessments may be relevant, and aspects of AI evaluation that remain insufficiently addressed.


Communication Preferences

The following options are entirely voluntary and independent of your survey participation. You may select any combination or leave all options unchecked. You can withdraw your preferences at any time by contacting GRDprocess at grdprocess@gmail.com


Contact Information

Providing your name and organization is optional. Your email address is required only if you have selected at least one communication preference above.

Any contact information you provide will be associated with your survey responses.


Privacy

Your responses will be processed by GRDprocess Sàrl in accordance with our Survey Privacy Notice.

If you provide contact information, it will be associated with your survey responses. You will only be contacted for the purposes you explicitly select.

To request access, correction, or deletion of your data, or to withdraw a communication preference, contact grdprocess@gmail.com.