人工智能安全治理框架-AI SAFETY GOVERNANCE FRAMEWORK.docxVIP

Content

1.PrinciplesforAIsafetygovernance 1

2.FrameworkforAIsafetygovernance 4

3.ClassificationofAIsafetyrisks 5

3.1AIsinherentsafetyrisks 6

3.2SafetyrisksinAIapplications 9

4.Technologicalmeasurestoaddressrisks 13

4.1AddressingAI’sinherentsafetyrisks 13

4.2AddressingsafetyrisksinAIapplications 15

5.Comprehensivegovernancemeasures 17

6.SafetyguidelinesforAIdevelopmentandapplication 21

6.1Safetyguidelinesformodelalgorithmdevelopers 21

6.2SafetyguidelinesforAIserviceproviders 23

6.3Safetyguidelinesforusersinkeyareas 24

6.4Safetyguidelinesforgeneralusers 26

-1-

AISafetyGovernanceFramework

AISafetyGovernanceFramework

(V1.0)

ArtificialIntelligence(AI),anewareaofhumandevelopment,presentssignificantopportunitiestotheworldwhileposingvariousrisksandchallenges.Upholdingapeople-centeredapproachandadheringtotheprincipleofdevelopingAIforgood,thisframeworkhasbeenformulatedtoimplementtheGlobalAIGovernanceInitiativeandpromoteconsensusandcoordinatedeffortsonAIsafetygovernanceamonggovernments,internationalorganizations,companies,researchinstitutes,civilorganizations,andindividuals,aimingtoeffectivelypreventanddefuseAIsafetyrisks.

1.PrinciplesforAIsafetygovernance

-Committoavisionofcommon,comprehensive,cooperative,andsustainablesecuritywhileputtingequalemphasisondevelopmentandsecurity

-PrioritizetheinnovativedevelopmentofAI

-TakeeffectivelypreventinganddefusingAIsafetyrisksasthestartingpointandultimategoal

-2-

AISafetyGovernanceFramework

-Establishgovernancemechanismsthatengageallstakeholders,integratetechnologyandmanagement,andensurecoordinatedeffortsandcollaborationamongthem

-EnsurethatallpartiesinvolvedfullyshouldertheirresponsibilitiesforAIsafet