AI model. Some of the secondary impacts resulting from a biased or erroneous AI model may not be remediable. Likelihood and Management Factors: The likelihood of impacts related to responsible AI and 23 their management may change based on different characteristics , such as: • Volume and sensitivity of data processed: If a SaaS service is inputting sensitive user data into its AI models, the likelihood of human rights harm is higher compared to a SaaS service that inputs machine data, for example. • Level of automation involved: If a SaaS service incorporates AI and ML models, the likelihood and severity of human rights harm may be higher. It is important to note that potential human rights impacts are related to how a customer uses the outputs of an AI model. However, the design of the AI model and the guidance the SaaS provider offers its customers are critical to those outcomes. Content-Related Issues Some SaaS services include platforms that are open to user-generated content or marketplaces where they offer integration with third-party applications. These functionalities may surface new human rights impacts related to content moderation. Company-Level Impacts • User-generated content posted on SaaS services may discriminate against certain individuals or populations. Online hate speech may also lead to offline violence, impacting individuals’ rights to life, liberty, and security. • SaaS providers operate in countries with varying views on their citizens’ rights to free expres- sion and access to information. Governments may ask SaaS providers to remove content from their platforms, limiting individuals’ access to information. These risks may change depending on datacenter locations and data localization regulations. • Third-party applications offered through the marketplaces of SaaS services may be used in ways that lead to adverse human rights impacts. Examples of Company-Level Impacts SaaS Provider One: A CRM solution offers integration with third-party applications through its marketplace. The applications include a tool that allows employees to leave private comments about clients, visible only to their coworkers, encompassing both work-related and personal characteristics. 26 Human Rights Assessment of the Software-as-a-Service Sector

Human Rights Assessment of the Software-as-a-Service Sector - Page 27 Human Rights Assessment of the Software-as-a-Service Sector Page 26 Page 28