This document describes Validators' policy regarding the development, procurement, and use of Artificial Intelligence (AI).
At Validators, we embrace the integration and developments of and around AI. Our main objectives are:
However, technology is not a solution for misalignment. AI can exacerbate this problem when used incorrectly and also make it less visible. A solid, measurable strategy that aligns every division, department, and individual is therefore essential.
The goal is to ensure the responsible, ethical, and safe use of AI in line with our corporate values, the interests of our customers and employees, and applicable laws and regulations. In addition, the sense of responsibility demonstrably diminishes when the application is no longer for individual use but is acted upon in the interests of the company.
Document and communicate when and how AI is used, and how AI contributes to specific tasks or decisions.
The use of AI should not replace ultimate human responsibility and liability.
The use of AI must be lawful and respect the rights of third parties and the Validators' Code of Conduct.
AI should be used in a way that promotes fairness and inclusivity and prevents bias.
AI must be adequately protected to limit cybersecurity incidents and ensure its safe use.
The use of AI must be evaluated with the aim of limiting ecological and societal risks.
The use of AI must remain under human control at all times and be critically assessed.
AI-generated output must be checked at all times. Responsibility lies at the individual level.
AI must be used in a manner that respects privacy and the protection of personal data.
In addition, Validators has an extensive awareness program that also focuses heavily on AI. The aim of this program is to transfer knowledge, raise awareness, and align views, opportunities, and obligations surrounding the use of AI. Participation in the program is NOT optional.
A multidisciplinary AI Committee will be established within Validators. This committee will have final responsibility for the implementation and enforcement of this policy. All proposals for the development, procurement, or significant modification of an AI system must be submitted to the AI Committee for approval before the project can commence.
The tasks of the AI Commission include:
All employees who deal directly or indirectly with AI systems in their work are required to participate in an AI awareness program. This program is essential for increasing AI literacy within the organization.
The program includes:
To guarantee data security and sovereignty, it is strictly required that all data processed by AI models and systems be stored and processed exclusively within the borders of the European Union (EU) or countries with an adequacy decision.
Companies that provide AI services or products to Validators must meet the following requirements:
Full compliance with the General Data Protection Regulation (GDPR).
Valid ISO 27001 (information security) and/or SOC 2 certification.
These requirements are laid down in a contract and compliance with them is periodically evaluated by the AI Committee.
The use of AI must offer demonstrable added value. Prior to implementation, a thorough analysis is made of the advantages (such as efficiency or new insights) and disadvantages (such as risks of errors or impact on employees). The use of AI must never be at the expense of the in-depth internal knowledge and expertise that forms the core of our services. AI is a tool that supports human judgment, not replaces it.
All AI systems are extensively tested and validated before they are put into operational use. We apply a risk-based approach that is aligned with the EU AI Act:
AI systems for internal process optimization with minimal impact. These undergo standard functional testing and quality control and do not have access to company, customer, or personal data. Users must be informed that they are interacting with an AI system.
AI systems with limited external impact, such as chatbots. Additional tests on transparency, reliability, and accuracy are mandatory for these systems. Users must be informed that they are interacting with an AI system. Human supervision is mandatory at all times.
AI systems that play a significant role in our services or in critical decision-making processes. In addition, systems where the model has access to company, customer, and/or personal data. These systems undergo a rigorous audit in terms of quality, robustness, security, fairness, and the prevention of bias. Human supervision is mandatory at all times.
To guarantee continuity and long-term use (many years), we strive for maximum independence from external parties. We achieve this by:
Validators recognizes the environmental footprint of AI technology. That is why we are taking the following measures:
We are transparent about the use of AI to employees and customers. Where AI is used in communication or service provision, this is explicitly stated. If we use respondent data to train models, respondents must be informed of this.
Critical decisions are never fully automated. For all high-risk systems, there is meaningful human oversight, whereby a human can override the decision proposed by AI.
We take active steps to prevent our AI systems from adopting or reinforcing existing biases. The datasets and algorithms are specifically checked for bias, particularly in high-risk systems.
Do you have questions about how Validators uses AI, or would you like to know more about our approach? Feel free to contact us.