SAP Business AI follows our core message: Relevant, Reliable and Responsible.
In this section we will cover the Responsible portion of SAP’s core message and focus on AI Ethics, Legal and Security topics.
Every AI system that we create at SAP must be based on ethical principles and values. We have moved from our own AI principles to follow UNESCO’s recommendation on the ethics of AI. These principles are very similar to the original SAP’s focus, but now follow a global standard for the peace of mind for our customers.
When building AI solutions, we must also respect applicable laws and regulations including the EU AI Act or similar laws in other regions and countries as they become available. Finally, we must ensure state of the art technical robustness from a data security perspective.
These three principles define guardrails for what we do:
- Ethical principles
- Respecting laws and regulations and
- Maintaining state of the art technical security
These principles guide and define our SAP Global AI Ethics policy which is mandatory for everyone at SAP that is building or involved in the creation of AI systems and solutions. An SAP AI Ethics Steering Committee also assists in deciding if use cases can be built or if potential solutions are not within our guidelines.
We also leverage external guidance in how we operate AI with an AI Ethics Advisory Panel consisting of sophisticated experts in the field who review our practices on an annual basis to ensure we are leading the way safely and securely.
SAP’s Guiding Principles for Ethical AI and Global AI Ethics Policy are available here and regularly updated in addition to the SAP AI Ethics handbook for applying SAP’s Global AI Ethics Policy.
So, what are the areas covered under Responsible AI? It includes:
- Human Agency and Oversight
- Privacy and Data Governance
- Fairness and Non-Discrimination
- Technical Robustness and Safety
- Transparency and Explainability
- And Society and Environment
But what areas of Responsible AI are represented in SAP’s AI Ethics Policy?
- Human Agency & Oversight: Safeguarding Human autonomy, particularly for automated decision-making. This also ties into ensuring that users can understand and influence what AI is doing.
- Fairness and Non-Discrimination: Patterns of marginalization, inequality, and discrimination must not be encoded into AI. We must also ensure that bias is removed from our AI outcomes.
- Transparency and Explainability: Prioritizing both the transparency of the development process and an AI’s decision.
Each of these policies is extremely important in ensuring that SAP remains responsible with our AI solutions.
Additionally, a detailed AI Risk Assessment and Classification process has been created. Using this process when new AI use cases are created, an Impact Assessment form must be completed. This assists SAP in ensuring that no use cases affect personal freedom, negatively affect society, or the environment. If any of these conditions (and others) cause a "Red Line" use case, the solution cannot be created as specified in the AI Ethics Policy.
Additional criteria have been defined such as:
- If the use case is processing personal data
- If the use case is processing sensitive personal data
- If the use case has the potential to negatively impact individuals
- If the use case contains automated decision making
- If the use case is in the high-risk sector
If the use case meets two or more of these criteria, it is defined as a "high-risk" use case and an Ethics Policy Self-Assessment must be created and presented to the SAP AI Ethics Steering Committee for additional assessment on the risk and potential mitigation. If less than two of the criteria affect the use case, the use case is defined as a standard use case and the Ethics Policy and only an Ethics Policy Self-Assessment is documented.
For additional details on AI ethics at SAP in addition to the resources linked above, please see the Putting AI Ethics into Practice at SAPlearning journey.
