Achieving Ethical AI: A Unified Vision

SAP's Commitment to Trustworthy AI

At SAP, the alignment of belief in the substantial potential of AI to create opportunities is evident through their dedication to 'Trustworthy AI,' a fundamental pillar within the company's strategy. This commitment is fortified by the establishment of a robust governance structure, providing guidance for the ethical and transparent development of AI. SAP's approach to AI Ethics is marked by a pledge to uphold the Universal Declaration of Human Rights, actively prohibiting discrimination and harassment. The core focus of 'Trustworthy AI' at SAP is a commitment to safeguarding customers and users, placing emphasis on fair and just treatment for all individuals. The SAP AI Global Ethics policy stands as the foundational framework, ensuring that the development of AI systems aligns with established principles and core organizational values. Additionally, the SAP AI Ethics Handbook meticulously directs SAP employees throughout the entire AI development lifecycle, mapping policy requirements to the respective stages of SAP's AI Factory processes.

Importance of interdisciplinary collaboration

SAP's commitment to trustworthy AI is reinforced through strategic collaborations with renowned academic institutions. By participating in the Corporate Affiliate Program of the Stanford Institute for Human-Centered Artificial Intelligence (HAI), SAP contributes to research, policy development, and education for responsible technology development based on artificial intelligence (AI). SAP supports HAI with expertise and resources, exploring the boundaries of generative AI in the corporate realm and understanding its transformative impacts on industries and economies. The partnership underscores SAP's dedication to ensuring that AI aligns with the principles of being relevant, reliable, and responsible, especially in its rapid evolution.

Additionally, SAP collaborates with the Sky Computing Lab at UC Berkeley, aiming to leverage cutting-edge research on interconnected cloud technology and accelerate its AI strategy for building hundreds of applications. This collaboration not only enhances SAP's understanding of large language model applications and general AI workloads but also contributes to reducing cloud costs and optimizing efficiency across different clouds.

SAP's Approach to AI System Differentiation and Risk Mitigation

Within SAP, a clear differentiation is made between Rule-Based AI and Learning-Based AI, substantiated by Risk Assessments conducted specifically on the latter during development phases. Rule-Based AI systems exhibit behavior entirely defined by human-created rules, while Learning-Based AI systems autonomously learn behavior, rules, and relationships. These Risk Assessments are systematically conducted to ensure the responsible and unbiased development of AI, aligning seamlessly with SAP's overarching commitment to ethical and transparent practices.

Fostering Responsible AI Practices at SAP: Key Principles

1. Human Intervention: At SAP, a foundational principle centers on prioritizing human intervention in all automated decision processes. This commitment emphasizes the integral role of human oversight, ensuring that decision-making incorporates safeguards against potential inaccuracies that may compromise the rights of data subjects. This principle underscores SAP's dedication to ethical AI deployment and responsible data processing.
2. Transparency in AI Usage: Transparency stands as a cornerstone in SAP's AI practices, particularly concerning personal data processing. Open communication about how AI is utilized during the handling of personal information reflects SAP's commitment to building trust with users and stakeholders. This dedication is evident in providing clear insights into the AI-driven processes impacting their data.
3. Bias Mitigation and Safeguards: SAP places a paramount concern on guarding against biases and inaccuracies in AI applications. The commitment involves implementing AI methods that prevent detrimental inaccuracies affecting individuals' rights. Additionally, SAP diligently implements safeguards to protect the personal data used in training AI models, ensuring the highest standards of privacy and security.

SAP's Approach to High-Risk AI Use Cases: A Rigorous Evaluation Process

SAP objectively classifies AI use cases as high-risk under specific circumstances, guided by a meticulous evaluation process. Criteria for this classification are based on past instances where AI has led to negative consequences for individuals or populations. High-risk cases undergo a thorough assessment by SAP's AI Ethics Steering Committee before advancing to further development, deployment, or sale. Criteria, including processing personal and sensitive data, automated decision-making, potential negative impacts on individuals, and domain categorization, guide this process, ensuring responsible AI applications across various domains.

Human-Centered AI: Anchoring Development in User and Stakeholder Needs

In SAP's commitment to responsible AI, a strong emphasis is placed on "Human-Centered AI." The belief in understanding and prioritizing user and stakeholder needs enhances the benefits of AI applications. Integral methods such as user research, design thinking, and defining user stories guide development processes. This approach ensures that SAP's products closely align with the real needs of target groups, actively avoiding unconscious bias and mitigating unintended harms. Through this commitment, SAP strives to build user trust, reduce fears, and contribute positively to society.

Conclusion:

SAP, with its commitment to Trustworthy AI, sets a benchmark for responsible AI practices. The governance structure, guided by principles rooted in transparency, human intervention, and bias mitigation, reflects SAP's dedication to ensuring fair and just treatment for all individuals. The meticulous evaluation process for high-risk AI use cases further demonstrates SAP's commitment to ethical and transparent practices, safeguarding against potential negative consequences.

Learn more about SAP’s approach to ethical AI:

• SAP AI ethics overview
• SAP’s Guiding Principles for AI
• SAP AI Global Ethics policy
• SAP AI Ethics Handbook
• openSAP training “AI Ethics at SAP
• openSAP training “Generative AI at SAP”