AIEI Membership Questionnaire

Introduction

Membership IDXXXXXX (Please enter the Member ID exactly as it appears on your membership certificate)

Prefix

First Name

Middle Name

Last Name

Email Address

Phone

Section 1: AI Ethics Governance & Accountability

This section examines your organization’s overall governance of AI ethics – the structures, policies, and leadership engagement that ensure accountability for ethical AI.

Q1. AI Ethics Leadership and Structure: Does your organization have an established AI ethics governance structure?Yes.No.Not Relevant.For example, an AI ethics committee, a responsible AI officer, or a dedicated task force.

Briefly describe who is part of this structure (e.g., a cross-functional ethics board, a single executive lead) and its role (e.g., reviewing AI projects, setting policies)

Explain how AI ethics oversight is currently managed (e.g., handled ad-hoc by project teams or part of general compliance)

Briefly explain why?

Q2. Policies and Frameworks: Do you follow a formal AI ethics policy or framework within your organization?No formal policy yet – we address issues case-by-case.In development – we are currently drafting or planning an AI ethics policy/guidelines.Yes, internal policy – we have an official internal AI ethics policy or code of conduct for AI.Yes, external frameworks – we adhere to external AI ethics guidelines/frameworks (e.g., OECD, AIEI, EU AI Act, IBM, Microsoft, Apple, etc.)OtherThis could be an internal policy/guidelines or adherence to external frameworks. Multiple Choice – select the option that best applies.

Please specify

Q3. Executive Oversight: How actively do senior leaders (executives or board members) oversee AI ethics in your organization?

Answer format: Rating Scale (1–5) – Rate the level of leadership engagement, where 1 = little to no executive oversight of AI ethics, and 5 = regular executive/board review of AI ethics issues (i.e., it’s a standing agenda item with defined accountability).

Optional elaboration: You may mention any specific executive roles (e.g., a CEO, CTO, CFO etc.) or departments responsible for AI governance, or note if leadership is updated on AI ethics risks periodically.

Q4. Accountability for AI Outcomes: Is there a clear process for handling accountability when an AI system causes an issue or harm?Yes.No.Not Relevant.In other words, have you defined who is responsible for AI outcomes and how to address negative impacts if they occur?

Describe what mechanisms are in place (e.g., an incident response plan for AI errors, an ethics review board that convenes if something goes wrong, or integration of AI risks into enterprise risk management)

Please discuss how such accountability is implicitly handled or any plans to formalize it (for example, “Not yet – but we plan to extend our risk management to cover AI-specific incidents”).

Briefly explain why?

Q5. AI Ethics Training & Awareness: Do you provide training or awareness programs on AI ethics for your employees?Yes.No.Not Relevant.

Briefly describe who receives this training and how often (e.g., “All developers and project managers take an annual AI ethics training; we also hold workshops for executives”)

Note if you are considering it or if your staff learn about AI ethics informally (e.g., through guidelines or on-the-job) or don't think it is nessesary.

Briefly explain why?

Section 2: Human Oversight & Human-Centric AI

This section explores how you ensure meaningful human oversight of AI systems, to maintain human agency and control over AI outcomes.

Q6. Human in the Loop: For which types of AI-driven decisions does your organization require human oversight or the possibility of human override?None / Fully Autonomous – Once deployed, our AI systems operate autonomously without human intervention.High-Stakes Only – Human review or override is required only for high-risk decisions (e.g. those affecting safety, legal rights, or critical customer outcomes).Most AI Decisions – In most cases, AI outputs are reviewed or can be adjusted by a human before finalizing the decision.Not Applicable – We currently do not use AI for autonomous decision-making.Multiple Choice – select the option that best applies.

Please give a brief example of how human oversight is implemented.For instance, “In a recent project automating loan approvals, staff had to sign off on any rejection that the AI recommended, ensuring a human final decision."

Q7. Reflective Prompt – Oversight in Practice: Think about a recent AI project in your organization. How was human oversight incorporated during that project’s deployment or usage?Answer format: Short Open-Ended – Describe one instance or practice. For example, was a human able to intervene or veto the AI’s decisions? Did you establish any rules like “if the AI’s confidence is low, escalate to a human”? If human oversight was not used in that project, explain why (e.g., the project was low-risk, or perhaps it revealed a gap in oversight that you plan to address). If the question is not relevant mention it.

Q8. Empowering Users/Staff: Do you provide tools or avenues for users or employees to flag issues or override AI decisions when appropriate?Yes.No.Not Relevant.

Briefly describe (e.g., “Employees using the AI forecasting tool can adjust predictions and provide feedback, which is logged for model improvements” or “Customers can appeal an automated decision via our support channel”)

Mention if you rely purely on AI outputs or if this question is not applicable to your AI use cases.

Section 3: Fairness & Bias Mitigation

This section looks at how your organization addresses bias, fairness, and non-discrimination in AI systems. Ensuring AI does not produce unfair or biased outcomes is a key principle of responsible AI.

Q9. Bias Identification Practices: Do you have a defined process to identify and mitigate bias in your AI models or datasets?Yes.No.Not Relevant.

Describe the process or tools you use (e.g., bias audits, fairness metrics in model evaluation, reviewing training data for representativeness).

Note any plans to implement bias mitigation, or explain challenges (e.g., lack of tools, not enough expertise, or you believe your use-case has minimal bias risk).

Briefly explain why?

Q10. Team Diversity & Stakeholder Input: How does your organization incorporate diversity or external perspectives to improve AI fairness?Diverse AI Team – We strive for diversity (in gender, ethnicity, background) within AI development teams to bring varied perspectives.User/Stakeholder Feedback – We solicit feedback from users or impacted groups to identify potential bias or fairness concerns.External Review – We engage external auditors or advisors (e.g., ethics consultants, academia, or third-party audits) to review our AI systems for fairness issues.None of the above – We do not explicitly do this yet.Other – (please specify).Answer format: Multiple Choice (select all that apply).

Please specify any other methods for ensuring fairness.

Q11. Fairness Testing & Metrics: How do you measure or evaluate fairness in your AI outcomes?Answer format: Short Open-Ended – Provide any tools, metrics, or tests you apply (if any). For example, you might mention using statistical parity or disparate impact analysis, running the model on subgroup data to check outcomes, conducting bias bounties, etc. If you do not currently test fairness, you can state that or describe how you ensure fairness in a less formal way (e.g., manual review of outputs for biases).

Q12. Self-Assessment of Fairness Maturity: In your view, how well has your organization addressed bias and fairness in AI so far?

Answer format: Rating Scale (1–5) – Rate your AI fairness maturity, where 1 = “We have not yet implemented formal bias mitigation” and 5 = “We have robust measures in place and regularly ensure our AI is fair and non-discriminatory.”

Optional elaboration: You may mention a brief example that justifies your rating.For instance, a 5 might note that you have a dedicated fairness toolkit and conduct bias reviews for every model; a 2 might note that you’re just beginning to explore fairness checks on one pilot project.

Section 4: Privacy & Data Governance

This section focuses on privacy protection and data governance in your AI activities. It covers how you handle personal or sensitive data used in AI in line with privacy principles and regulations.

Q13. Personal Data and Privacy Measures: What steps do you take to protect privacy in AI projects involving personal data?Compliance with Privacy Laws – We ensure AI systems comply with relevant regulations (e.g., GDPR, HIPAA) regarding personal data.Data Minimization/Anonymization – We anonymize data or minimize the personal data we use (only what is necessary for the AI’s purpose).Access Controls – We restrict access to sensitive data (only authorized team members can access training or output data).Privacy by Design – We implement technical measures like encryption, differential privacy, or federated learning to enhance privacy.User Consent and Notice – We obtain consent for data use and inform users how their data is used in AI systems.Other – (please specify).None specifically – We do not have special measures beyond standard IT security policies. (please elaborate).Answer format: Select all that apply (multiple selection).

Please specify any other privacy practices.

Please elaborate why or if this is an area for future improvement.

Q14. Privacy Impact Assessments: Do you conduct any form of Privacy Impact Assessment for AI initiatives?Yes.No.Not Relevant.Such as a Data Protection Impact Assessment – DPIA – if under GDPR, or an internal privacy review.

Please note at what stage and how (e.g., “Yes, we perform a DPIA at the start of any project involving personal data”)?

Mention if there’s an alternative approach (e.g., “No formal DPIA, but privacy is reviewed as part of our general project risk assessment”).

Briefly explain why?

Q15. Data Quality and Governance: How do you ensure data quality and integrity for your AI systems? (For instance, avoiding inaccurate or biased data which can affect outcomes.)Answer format: Short Open-Ended – In a few words, describe any data governance practices: e.g., data validation processes, routine audits of datasets, documentation like datasheets for datasets, or roles like data stewards. If you don’t have formal practices, you can note that (e.g., “We rely on data collected during operations and haven’t established formal governance yet”).

Section 5: Transparency & Explainability

This section covers transparency (being open about AI use) and explainability (ability to explain AI decisions). These practices build trust and enable understanding of AI decisions by stakeholders.

Q16. AI Usage Disclosure: Do you inform users or affected individuals when AI is involved in making a decision or providing a service?Yes.No.Not Relevant.

Describe how you communicate this?e.g., “Our app labels recommendations as ‘AI-generated’,” or “Customers are notified in our privacy policy or at point of interaction that an AI is being used”

Please explain why not?e.g., “Our AI is back-end only, users don’t interact with it directly” or “We have not developed a communication strategy for this yet”

Briefly explain why?

Q17. Explainability: Are you able to provide explanations for how your AI systems reach their decisions or outputs, especially to those who are impacted by them?Low explainability (Black Box) – We generally cannot interpret or explain the AI’s decisions in understandable terms.Basic explainability – We can provide some basic rationale behind decisions (e.g., key factors the AI considered), at least for certain systems.High explainability internally – We have tools or documentation (such as model interpretability tools, model cards) that allow us to explain AI outputs to an internal team or an expert audience.User-level explainability – We have mechanisms to explain AI decisions to users or customers (e.g., providing reason codes or an explanation interface on user request).Not applicable – Our AI systems are not the type that make decisions requiring explanation or the question is not relevant.Answer format: Multiple Choice – select the option that best describes your situation.

Please add a brief example.e.g., “For our AI credit scoring, we can show the top 3 factors influencing a particular score to loan officers and customers”

Q18. Internal Documentation & Traceability: Do you maintain internal documentation or logs for your AI systems to support transparency and accountability?Yes.No.Not Relevant.e.g., model cards, datasheets for datasets, audit logs of AI decisions

Briefly list what kinds of documentation?For example, “We create a ‘model fact sheet’ for each AI model detailing its intended use, limitations, and performance. We also log each automated decision for audit purposes”

Mention if this is under consideration or how you currently keep track of AI system behavior (even informally).

Q19. Scenario – Explaining AI Decisions: Imagine a customer or client asked why an AI made a specific decision about them (for example, being denied a service or receiving a certain recommendation). How would your organization handle this?Answer format: Short Open-Ended – Describe briefly the process you would follow. For instance, would your support team be able to retrieve an explanation from the AI system or from documentation? Is there a policy to provide such an explanation upon request? If this scenario has happened before, you can describe what was done. If it has not happened, outline what you would do given current capabilities (or note that you currently would not be able to provide a clear explanation).

Section 6: Security, Safety & Technical Robustness

This section examines how you ensure the security and technical robustness of your AI systems – that they are safe, reliable, and resistant to failures or misuse. Robust and secure AI is vital to prevent harm and maintain trust.

Q20. AI System Security: Do you take specific measures to safeguard your AI systems and data from cyber threats or unauthorized access?Yes.No.Not Relevant.

Provide examples of security measures.e.g., “We conduct regular penetration testing on our AI infrastructure,” “We have access controls and encryption for model and data storage,” “We monitor for adversarial inputs or data poisoning attempts”

Please explain if you rely on general IT security only or if this is a gap you plan to address (given that AI systems can have unique vulnerabilities)

Briefly explain why?

Q21. Reliability & Validation: How do you ensure your AI systems are technically robust and reliable in their performance?Answer format: Short Open-Ended – Describe any testing or validation practices (for example, testing on diverse scenarios, validating outputs with ground truth, simulation of edge cases, stress testing the model under different conditions). You can also mention if you track performance metrics post-deployment to catch degradations or errors. If an AI system makes a critical mistake, how would you catch it? (E.g., manual review of samples, automated alerts on unusual outputs.) If you do not have formal tests beyond typical software QA, note that.

Q22. Fallback and Human Intervention Plan: If an AI system fails, malfunctions, or produces incorrect/harmful results, is there a fallback plan or human intervention procedure?Yes.No.Not Relevant.

Briefly describe what is in place.e.g., “If our AI malfunctions, it automatically hands control back to a human operator,” or “We have an on-call team to shut down or fix the AI service if issues arise”

Mention if you implicitly rely on system design that “it won’t fail,” or if this is an area you need to develop (having a fallback or manual override is a safety best practice).

Briefly explain why?

Q23. Confidence in AI Safety: Overall, how confident are you in the safety and robustness of your AI systems? (This considers security against attacks, reliability of outcomes, and having safeguards for failures.)

Answer format: Rating Scale (1–5) – 1 = Not confident: we have significant unaddressed risks or haven’t assessed this; 5 = Highly confident: we have strong assurances through testing, safeguards, and security controls.

Optional elaboration: You may comment on any particular reason for your confidence level or concerns.e.g., “We rated 2 because while we secure data, we haven’t tested against adversarial attacks,” or “We rated 5 due to rigorous testing and an ISO 27001 certified pipeline”.

Q24. Final Question – Help Us ImproveDo you have any feedback or suggestions on how we could improve this questionnaire? (For example: Were any questions unclear, not relevant to your organization, too detailed, or missing something important?). Your thoughts are greatly appreciated!

Declaration

The information you provide in this questionnaire will be used solely for research purposes to help AIEI enhance the support, resources, and benchmarking insights offered to our members. Additionally, aggregated and fully anonymized findings may be shared with governmental and industry stakeholders to promote broader understanding, awareness, and informed discussion on ethical AI practices. We emphasize that your responses will remain completely confidential and securely stored, presented only in a collective form where no individual or organization can be identified. Your participation is entirely voluntary, and your decisions regarding questionnaire completion will not affect your organization's membership status, reputation, or relationship with AIEI in any way. We fully respect your choices, and any recommendations provided based on the survey results are purely optional for your consideration.

Save as Draft

Instructions:

AI Horizon Conference