AI Content Chat (Beta) logo

SHAREHOLDERPROPOSALS groups have used Amazon Rekognition’s facial recognition capabilities to protect human rights, including tracking and stopping child exploitation and rescuing victims of human trafficking, as well as locating hundreds of missing children. It has also been used to build educational apps, enhance security through multi-factor authentication, identify suggestive or explicit website content in order to block or remove those images, and provide identity verification as part of mobile banking services for underbanked individuals in emerging geographies. These are just a few of the numerous beneficial applications of the technology. WeAreCommittedtotheResponsibleUseofOurAI/MlProductsandServicesandHaveTakenNumerousActionstoAddress Concerns Around Potential Misuse of Rekognition Since introducing Amazon Rekognition, we have been consistent and proactive in our efforts to address concerns and mitigatetheriskofmisusethroughpolicyandadvocacyefforts,customercontractualrequirementsandtraining,consultation with third party experts, and other policies and practices. We understand the risks associated with potential misuse of facial recognition technology and, in connection with extensive discussions with customers, researchers, academics, policymakers, and civil society groups, we have taken the following actions to review and address concerns around potential misuse: • Implemented Police Moratorium. In June 2020, AWS implemented a moratorium on use of Amazon Rekognition’s face comparison feature by police departments in connection with criminal investigations and, in May 2021, AWS announced the indefinite extension of that moratorium. We believe this moratorium will give governments time to implement appropriate rules, and we stand ready to help with any such initiatives. Since this announcement, several United States stateandlocaljurisdictionshaveintroduced,debated,andimplementedsuchlaws,andweanticipateadditionalactivityand progress in this area. We support the calls for an appropriate national legislative framework that protects individual civil rights and ensures that governments are transparent in their use of facial recognition technology, and have provided 107 guidance to those thinking about these issues. • Actively Engage in Policy Discussions. Amazon believes that facial recognition technology should not be banned or condemnedsimplybecausethereisapotentialthatpeoplemaymisuseit.Manytechnologies,likecellphonesorcameras, could also be misused. Instead, as we have made clear in our statement of positions, “we think that governments and lawmakersshouldacttoregulatetheuseofthistechnologytoensureit’susedappropriately,andwehaveproposed guidelines for effective regulatory frameworks and guardrails that protect individual civil rights and ensures that 108 In addition to our implementation of the governments are transparent in their application of the technology.” moratorium on police use and legal terms for law enforcement use, AWS continues to engage with a large number of diverse stakeholders on these issues, including civil society groups, academia, policymakers, and law enforcement officials. • Dedicate Significant Resources to Machine Learning Accuracy and Bias Mitigation. AWS dedicates significant resources to testing, auditing, and improving its technology so that it is constantly learning and improving accuracy, including providing diverse perspectives on its technology development teams, using training data sets that reflect gender, racial, ethnic, religious, and cultural diversity, and incorporating feedback from third parties. We have science and technical experts who help promote fairness in our products and services, including helping to design, test, and audit our services for fairness andaccuracy and to mitigate potential bias, and who publish academic papers and provide thought leadership in this area.109 AWS also makes available capabilities that help customers detect bias in ML models and increase transparency 110 byhelping explain model behavior to stakeholders and customers. Wecontinuetoinvestheavily in this area and work closely with customers and other stakeholders on addressing these important issues. • Support Standardized Testing Methodologies and Benchmarks. We believe it is important that there be standardized testing methodologies and benchmarks for cloud-based facial recognition technologies. AWS encourages and supports the development of independent standards by entities like the National Institute of Standards and Technology (NIST) and other independent and recognized research organizations and standards bodies to develop tests that support cloud- basedfacialrecognitionsoftware.WeareengagingwithNISTandotherstakeholderstoofferourdirectassistancetowards this effort. We also support efforts by members of the academic community to establish independent and trusted criteria, benchmarks, and evaluation protocols around facial recognition services. 107 Available at https://aws.amazon.com/blogs/machine-learning/some-thoughts-on-facial-recognition-legislation/. 108 Available at https://www.aboutamazon.com/about-us/our-positions and https://aws.amazon.com/blogs/machine-learning/some-thoughts-on-facial- recognition-legislation/. 109 Available at https://arxiv.org/abs/2007.06570 and https://www.youtube.com/watch?v=JCGUYFe6P2k. 110 Available at https://aws.amazon.com/sagemaker/clarify/. 84

Amazon 2022 Proxy Statement - Page 92 Amazon 2022 Proxy Statement Page 91 Page 93