AI Policy

Purpose

This document serves as Feedzai – Consultadoria e Inovação Tecnológica, S.A. and all its group companies (“Feedzai” or “we”) Corporate AI Policy, outlining the company’s steadfast commitment to pioneering responsible and ethical artificial intelligence (AI) practices in the field of financial risk management.

The purpose is to articulate Feedzai’s approach to developing, deploying, and maintaining AI solutions that not only showcase innovation but also adhere to the highest ethical standards and global compliance norms. By detailing Feedzai’s contributions to Responsible AI, ethical principles governing AI development, and client engagement initiatives, this policy aims to provide comprehensive insights into the company’s dedication to transparency, fairness, privacy, accountability, and ethical design in the realm of AI. Ultimately, this policy underscores Feedzai’s leadership role in the industry and its commitment to advancing responsible AI practices in financial risk management.

This Policy applies to all of our directors, officers, employees and any other associated persons of Feedzai (such as contractors). We refer to them collectively as “Feedzaians” unless the context otherwise requires.

Pioneering Responsible AI

Feedzai has established itself as a leader in the field of Responsible AI. Our team’s innovative contributions have enriched the open-source community, setting new benchmarks for ethical AI development.

Our key developments, including Fairband, an acclaimed tool for fairness optimization, and FairGBM, which integrates fairness constraints directly into model training, underscore our commitment to advancing the field. Furthermore, our development of TimeSHAP enhances the explainability in deep learning, and our rigorous Bias Audits using Aequitas ensure equity and fairness are at the forefront of all AI applications.

AI Ethical Principles

Our principles of AI ethics are the cornerstone of our operation. Transparency is paramount; we ensure that our AI processes are clear and that our clients fully understand how decisions are made.

In our pursuit of fairness, our product allows us to measure model bias and to train models with less bias. Upholding data confidentiality and integrity is non-negotiable, and we implement strict measures to protect sensitive information.

Moreover, accountability is ingrained in our ethos; we champion human oversight and governance of our financial risk management solutions.

AI Development Practices

Our approach to AI development is rooted in robust governance. As part of our product development and offering we provide a comprehensive suite of tools for the governance and reporting of machine learning models, covering the whole process, from data pipelines and transformations to feature development.

Our rigorous testing protocols include detailed bias audits and robustness testing, enabling our data scientists to adhere to the strict governance policies as defined by our clients. The ethical design and deployment of our AI solutions are central to our practice, focusing on fair and transparent decision-making processes.

Furthermore, our commitment to explainable AI can produce explainable models and model decisions, and, therefore allow compliance with applicable regulatory requirements.

Compliance, Security, and Legal Considerations

We are vigilant in ensuring that our practices align with existing laws and regulations, such as the General Data Protection Regulation (GDPR), and will prioritize the privacy, security, and data protection of individuals when designing and using AI systems.

We are vigilant in ensuring that our practices align with forthcoming regulations, including the EU AI Act.

We follow industry-standard security best practices, as evidenced by our PCI DSS Level 1 and ISO/IEC 27001 certifications, and our SOC 2 report. Our Information Security Management System (ISMS) includes our AI practices in its scope, which ensures an acceptable level of security in its development and use.

Furthermore, we align model development with best practices, such as existing guidelines, e.g. “Guidelines for secure AI system development”, developed by the U.S. Cybersecurity and Infrastructure Security Agency (CISA) and the UK National Cyber Security Centre (NCSC). These allow us to ensure secure AI design, development and operation.

Our AI solutions undergo regular audits and assessments to guarantee ethical integrity, data provenance, and legal compliance.

Environmental and Social Sustainability

Feedzai will strive to develop and use AI technologies in a manner that promotes environmental and social sustainability, minimizing negative impacts on the environment and communities.

Client Engagement and Education

Our relationship with clients is founded on collaboration and open communication. We use client feedback as a catalyst for continuous improvement in our AI solutions.

This collaborative ethos extends to our educational initiatives, where we provide resources and support to educate our clients on responsible AI use and the capabilities of our AI products.

By fostering this level of engagement, we ensure that our clients are not only recipients of our technology but active participants in a shared journey towards ethical AI developments and usage.

Conclusions

Feedzai’s unwavering commitment to responsible AI is more than a policy – it is an integral part of our mission in advancing financial risk management. This Corporate AI Policy reflects not just our dedication to ethical AI practices but also our role as a leader in the industry.

By adhering to these principles and practices, we aim to not only innovate but also set the standard for responsibility and ethics in the ever-evolving landscape of artificial intelligence.

If you have any questions about these guidelines, please contact the Research team at [email protected].

Enforcement

Violations of this policy may result in disciplinary action, up to and including termination of employment or contractual relationship.

Policy Review

This Policy will be reviewed annually or as needed, based on the evolution of AI technology and the regulatory landscape.