大型组织AI安全挑战应对策略英文版.pptxVIP

AICyberBenchmark

Howarelargeorganizationstackling

theAISecuritychallenge?

©WAVESTONE|2

Nodoubt:AIisa

uniqueopportunity…

…thatmustbesecured!

WavestoneAICyberBenchmark-2025

AIsystemsworkdifferentlyfromclassicITsystems…

INPUTS

Ontheinputs…

OUTPUTS

Ontheoutputs…

©WAVESTONE|3

WavestoneAICyberBenchmark-2025

…andcanthereforebeattackedinveryspecificandnewways

EVASION

ORACLE

POISONING

CHATGPTTRAINING

DATASETLEAK

MITHRILSECURITY

POISONINGTEST

AUTONOMOUSCAR

WavestoneAICyberBenchmark-2025

Regulatoryapproachesacrossgeographies

•Risk-basedapproach.

•EveryorganizationmustcomplybyMay2027.

•Alreadysomeconsequences:newiPhonewithGenAIChatGPTvoicechatfunctionalitypospotned…

•Chinaisfocusingonthe

cybersecurityofitssystemwitharisk-basedapproachandonregulatingtheprocessingofdata,especiallylabeling.

•RescindformerExecutiveOrder

14110thatprovidedguidelines.

•AimstoremoveanypotentialbarrierstoAIdevelopment.

varysignificantly

EUROPE

CybersecurityrequirementsforGenAIservices

InplacesinceMay2024

ExecutiveOrder14179

InplacesinceJanuary2025

AIACT

InplacesinceMarch2024

Ourgoal:helpclarifyhowtotackletheAIsecuritytopic,troughourAICyberBenchmark

SomeclientsareadoptingAIonalargescale:

•Between50and400usescasesidentified

•AstrongmobilizationatExcomlevel

WebenchedtheseclientsontheirAImaturity,basedonthe5NIST’spillars,andconsolidatedthoseresultstoproducethisfirstAICyberBenchmark.

WavestoneAICyberBenchmark-2025

Today,beyondthehypeeffect,AIisareality!

Workedwith+20

clientsalready

workingonthetopic.

•Govern

•Identify

•Protect

•Detect

•Respond

Leadingtoalotofactivitiesbutalotofblurriness.

©WAVESTONE|6

•Buildandsometimessell