2. Volume and sensitivity of data processed SaaS services rely on data input on varying levels. The degree to which a service processes sensitive personal data may surface different kinds of human rights impacts. SENSITIVE Software using non-sensitive data Software using sensitive data NON-SENSITIVE• Bias and discrimination can still occur, • The likelihood of human rights harm but in less visible ways. For example, related to privacy and data governance if a factory operations software is rises as the volume and sensitivity of the designed for a certain market, it may data processed by the SaaS provider not consider differences in the ways increases. Depending on where the that factories operate in other geogra- data is stored, SaaS providers may be phies. connected to potential human rights • Even if sensitive personal data is not harms on various levels and may or may used, algorithms may have unintended not have leverage to address them. ripple effects on human rights. For • If a SaaS service is inputting sensitive example, using seemingly benign user data into AI models, the likelihood machine data, companies may make of human rights harm increases. decisions about labor scheduling and worker conditions. 3. Level of automation involved Some SaaS services incorporate machine learning and artificial intelligence systems to analyze data and generate insights. The degree to which a SaaS service involves automation may lead to different kinds of human rights impacts, or it may exacerbate the severity or likelihood of certain risks. TED MORE AUTOMA Software involving less automation Software involving more automation • There is increased risk of human error • When a SaaS service incorporates AI LESS AUTOMAleading to human rights impacts. models and automation, the likelihood TED and severity of human rights risks may be higher. • The use of AI may also lead to an increased number of errors going unno- ticed, making human rights impacts less easily remediable. • SaaS providers may be more directly linked to harms caused by flawed algo- rithms, but they may also have more leverage to remediate adverse impacts. 46 Human Rights Assessment of the Software-as-a-Service Sector