<p>about your learning experience to help us improve.</p>

You've completed

cleanFiles/bookmap_Putting_AI_Ethics_into_Practice_at_SAP_en-US_learning-sap-com_lsc-prod.zip (MMCP-DROPZONE) - 2024-11-07T08:50:40.000Z

Governing AI Ethics at SAP: From Policy to Practice

Lesson Content

Putting AI Ethics into Practice at SAP

Explore how SAP governs AI ethics from policy to practice through a structured governance framework involving executive board oversight, steering committees, ethical assessments, and red line criteria.<br><br># This description was auto-generated #

SAP_MOOC_aie2_Lesson_3_GOVERN_Video.mp4

Hello, my name is Vikram Nagendra, and I drive the AI ethics governance at SAP within corporate sustainability. I'll present to you the lesson governing AI ethics at SAP from policy to practice. We have established the following governance structure at SAP to oversee ethical practices of AI. At the highest level is the executive board with the CEO as the board sponsor for AI ethics. The executive board sponsors the overall AI program, approves ethical framework elements like the global AI ethics policy, and has the final accountability of any risks emerging from high-risk AI use cases. The global AI ethics steering committee consists of senior leaders nominated by all executive board members plus senior leaders from relevant lines of businesses. The steering committee provides strategic direction and decisions within the company. Operating on behalf of the executive board, they also provide recommendations and approvals to continue the high-risk AI use cases. The AI ethics office centrally manages and orchestrates ethical AI work across the company. The governance aspects managed by the corporate sustainability team provides the overall ethical framework and orchestrates the process globally. Corporate Sustainability Team operates the governance, ensures independent oversight of the global framework, and maintains no conflict of interest with any outcomes. The operationalization aspects managed by the central AI unit implements the defined framework across the company and ensures that it is applied consistently. The external AI Ethics Advisory Panel, consisting of people in academia, business, and having policy-making experience, provides outside-end perspective and challenges SAP on its ethical journey. The panel also supports interviewing selected high-risk use cases that require a broader understanding of ethical implications. The slide shows the process that all AI work at SAP needs to go through before being developed, deployed, sold, or used. The starting point is triggered from a variety of sources like Sirius, which is an internal tool for orchestrating standard software development, or other tools. The first step is to complete the impact assessment, which classifies the use case into either a red line or not. In case of a red line case, further development is prohibited and the work needs to stop immediately. If it's not a red line, the next determination is whether the work should be considered high risk or not. This is done depending on a range of questions like whether the AI system processes personal and protected personal information, whether the work is in a high-risk sector, and so on. If it's classified as high-risk, then the AI Ethics Office will conduct a formal follow-up to understand the case better, and guide the teams to go through the Steering Committee for formal approval or rejection. Depending on the final decision from the Steering Committee, the teams will continue the work. If the classification is low-risk, then the teams can continue the work, including assessing their work against the policy using the self-assessment tool. Now let's take a closer look at what the red lines are. The first category of red lines relate to personal freedom. If the AI system is intended for targeting individuals or groups with the purpose of not respecting their fundamental rights, then it is classified as a red line under human surveillance. Another situation relates to discrimination of individuals or groups for specific opportunity. It's also considered a red line when anonymized data can be de-anonymized. The second category of red lines relate to civic society. If the AI system is meant for deception or unfair manipulation, or its use can result in harmful impacts, then it's prohibited. It is also prohibited for AI systems to undermine democratic systems and perform social scoring. The final category of red lines relate to the environment. If the AI system can have adverse impact on the environment, then it is also prohibited. Let's take a look at the conditions that would result in a high-risk classification. The first two conditions relate to processing of personal and protected personal data without anonymization. This could result in several risks like privacy breaches, bias and discrimination, and so on. When automated decisions are made without human oversight, the risk of propagating wrong decisions from inappropriate model states greatly increase. The fourth condition checks if the AI system could indirectly impact individuals or groups negatively. And finally, the fifth condition looks at high-impact industry sectors where the AI system is intended to be deployed. If an AI system meets one of these conditions, then a follow-up discussion with the AI ethics office would be conducted to determine if the system needs to be approved by the global Steering Committee. In conclusion, SAP is governing the ethical development of AI systems through various bodies like the Internal AI Ethics Steering Committee and the External AI Ethics Advisory Panel. Furthermore, SAP has established a risk-based AI ethics assessment process to guarantee that every use case is developed in accordance with the company's guiding principles and AI ethics policy. This demonstrates SAP's commitment to prioritizing ethical considerations in its AI development. With that, we have reached the end of our lesson. Thank you and goodbye.