2185 - 信任但要核实：持续的人工智能监控.pdf

《2185 - 信任但要核实：持续的人工智能监控.pdf》由会员分享，可在线阅读，更多相关《2185 - 信任但要核实：持续的人工智能监控.pdf（10页珍藏版）》请在三个皮匠报告上搜索。

1、October/25Intelligence:Transformation and Responsibility Trust but verify:continuous AI monitoring How is the journey of AI happening at BB?Trust as an Essential Asset Over 200 years of Banco do Brasil history Core asset:societys trust Trust protected and expanded by AI Governance Responsibility as

2、the guide for technology adoptionScaling AI withResponsibility Use of international frameworks(EY)Strategic partnership with IBM for operationalization Safe,ethical,and transparent innovationWe train teams,hold hackathons,and show that Responsible AI is not just IT orgovernanceit is everyones respon

3、sibility.Living CultureThe development of ethical AI is not a destination,but an ongoing journey.Continuous Commitment to EthicsWe start with simple agents,with no autonomy and no personal data.We only advance when governance is mature.Phased scalingEvery use case is registered,evaluated,and classif

4、iedMandatory cataloging of all AI agentsWe use market solutions to ensure quality,explainability,and compliance.Supervision tools and metrics:Protect personal data and ensure that AI operates safely.Active monitoringGovernance in Action at BBAsset CatalogingAsset ManagementF P Data QualityBestPracti

5、cesTrainingModelingDevelopmentComplianceRegulationData PrivacyData Protectionand PrivacyValidation andTestingCommitteesLines of DefenseApprovalsActorsHuman in The LoopImpactAssessmentImplementaBon PlanDeploymentRetrival AugmentedEvaluation LLM/Agent as a Judge LLM Evaluation Ground TruthContinuos Mo

6、nitoringRed Team Guard Rails Responsible AI AI GovernanceAddition to the pillars,we also structured the stages of the AI lifecycle,which help turn principles into practice:Pre-Cycle RequirementsLicenseApprovalLifecycle IAResponsible AI:Opportunities and RisksD