Telefónica´s Approach to the Responsible Use of AI 

Artificial Intelligence is rapidly growing and has various applications in domains such as content recommendations, chatbots, image recognition, fraud detection, medical diagnosis, and more. Telefónica uses AI and big data in core business optimization, customer relationship innovation, B2B services, and social good initiatives. However, concerns exist regarding discrimination, interpretability, and data transparency. To address these, Telefónica published AI Principles for responsible use in October 2018. 

The approach to the responsible use of AI includes a strategic model, an organisational and relational model, and an operational model, as detailed below. 

The strategic model: Telefónica’s Principles of Artificial Intelligence 

Telefónica is fully dedicated to respecting Human Rights, as stated in its Business Principles and Human Rights Policy. This includes developing products and services that create a positive impact on society and the environment. They believe in using technology, including AI, to foster inclusivity and provide better opportunities for all. Telefónica has published its “Principles of AI” to guide the responsible use of AI and Big Data across the company. The principles include: 

• Fair AI: Telefónica ensures that AI applications result in fair outcomes, free from discrimination based on personal characteristics. They prioritise considering the algorithm’s impact in specific domains, alongside error optimisation. 
• Transparent and Explainable AI: Involves disclosing data usage and purpose. Users should be aware of this when interacting with AI systems. Decision-making processes, especially in supervised machine learning, should be explained and tailored to the user’s transparency preferences. This applies to third-party AI technology as well. 
• Human-centric AI: Means that AI should serve society and provide tangible benefits to people. It should always remain under human control and guided by values. AI employed in products and services must not have any adverse effects on human rights or hinder the achievement of the UN’s Sustainable Development Goals. 
• Privacy and Security by Design: Integrates privacy and security aspects into the lifecycle of AI systems. This safeguards individuals’ right to privacy and personal data. Telefónica already has processes in place to ensure proper privacy and security measures, extending beyond AI systems. 
• Working with partners and third parties: Telefónica ensures that the logic and data used by providers align with its principles. These principles are based on expert consensus and specific considerations within the telecommunications industry. While practical implementation of such principles in large organizations is relatively limited, Telefónica’s approach represents a significant step towards responsible use of Artificial Intelligence. 

The organisational and relationship model 

We are implementing responsible AI by establishing an organizational and relationship model that outlines the involvement, roles, and interactions of different areas within the company. We encourage a self-responsibility approach with on-demand escalation. This includes a three-step escalation process, as depicted in the figure below. 

Product managers and developers using AI conduct a self-assessment during the design phase to identify human rights risks. They can seek assistance from the Responsible AI Champion. If concerns persist, multidisciplinary experts are consulted. Serious issues are escalated to the Responsible Business Office for resolution. 

The operating model 

The operating model implements the Responsible AI Approach on a daily basis, integrated with the broader Responsibility-by-Design-Approach. It includes the “Responsible AI by Design” methodology, influenced by Privacy and Security by Design. The operating model consists of, among others: 

Training & awareness 

Telefónica has developed accessible AI & Ethics courses for employees, available in three languages (Spanish, English, and Portuguese). Depending on the employee’s profile, they can access a shorter version divided into three 10-minute modules or a more comprehensive version that takes about an hour to complete. In addition to online courses, business units using AI participate in dedicated workshops to explain the governance model. An intranet mini-guide is also available for employees to quickly apply the AI Principles and ensure responsible AI usage. 

Questionnaire with recommendations 

For products and services using AI or Big Data, the responsible manager completes an online self-assessment questionnaire based on AI Principles. Completed questionnaires are logged for review, and managers receive an email with recommendations and unresolved issues. Collaboration across departments shaped the methodology, and privacy and security matters are handled separately by dedicated areas. 

Technical tools 

Our methodology utilizes both in-house and external tools to answer questionnaire questions. In-house tools include privacy-enhancing technologies and tools for detecting algorithmic discrimination and personal data transparency. We constantly evaluate new external tools, especially in the areas of bias/discrimination and explainability. The Community of Experts recommends the inclusion of new tools. 

You can find out more information about Telefónica AI principles at the following link: