Waaxaaradda Difaacu waxay si rasmi ah u amartay hoggaamiyayaasha sare ee ciidamada Maraykanka inay ka saaraan alaabada AI ee Anthropic nidaamkooda muddo 180 maalmood ah, sida ku cad warqad gudaha ah oo dhacday 6-dii Maarso oo CBS News helay. Amarkaas, oo la bixiyay maalintii ka dambeysay in Pentagonku Anthropic uga tago khatarta silsiladda sahayda, ayaa digaya in AI-ga Anthropic uu khatar aan la aqbali karin ku hayo dhammaan nidaamyada iyo shabakadaha Waaxda Difaaca. Warqadda oo ay saxiixday Madaxa IT-ga waaxda Difaaca Kirsten Davies, ayaa iftiimisay tallaabooyinka ballaaran ee talisyada ay tahay inay qaadaan si loo baabi'iyo AI-ga Anthropic ee nidaamyada muhiimka ah ee amniga qaranka, oo ay ku jiraan hubka Nukleerka, difaaca gantaallada, iyo dagaalka cyber-ka. Waxaa sidoo kale la amray in cid kasta oo qandaraaslaha ah ee la shaqaysa Pentagon ay joojiso isticmaalka alaabada Anthropic ee mashruucyada la xiriira Difaaca gudaha 180 maalmood. Davies ayaa ka digtay in hostaha cadowgu ay ka faa'iideysan karaan nuglaanta ku jirta hawlgallada Pentagonka, taas oo keeni karta khataro burbur leh oo ku habsada ciidammada. Waxay adkaysatay in oo kaliya ay adigu ansixin karto ka madax-bannaanida, kuwaas oo keliya la siin doono markasta oo ay ku jirto howlgal muhiim ah oo amniga qaranka aan lahayn beddelaadyo la heli karo, iyadoo ku xiran qorshayaasha la gudbiyay ee yareynta khatarta. Mas’uul sare oo ka tirsan Pentagon ayaa xaqiijiyay run ahaanshaha warqadda, halka Anthropic aysan wax faallo ah siinin waqtigan xaadirka ah. Tallaabadan aan hore loo arag ee dowlad federaal ah ayaa calaamadeysay markiisii ugu horreysay ee shirkad Maraykan ah loo qoro khatar silsiladda sahayda ah; xadidnaanta hore waxay ahayd kuwo shisheeye sida Huawei.

Tallaabadan ayaa raacday بی توافق – oo ay ku jirtay su’aalo ku saabsan codsiga Anthropic ee laba "xuduudood" oo mamnuucaya ciidanka Maraykanka inay isticmaalaan Claude AI si ay u sameeyaan kormeer ballaaran oo ku saabsan Maraykanka ama hubka si buuxda isu-matooyey—tallaabooyinka Dario Amodei, oo ah Maamulaha Guud ee Anthropic, ayaa sheegay inay yihiin kuwo muujinaya qiimaha Maraykanka. Laakiin Pentagonku wuxuu ku adkaaday isticmaalka Claude ujeeddooyinka sharciga ah dhammaan isticmaalka sharciga ah iyada oo aan xad lahayn, isagoo sheegaya in isticmaalka mamnuucay horey loo xadiday. Sida ay sheegeen ilo xog-ogaal ah, Claude waxay taageertaa dadaallada militariga Maraykanka ee dagaalka Iran. Anthropic ayaa ah shirkadda keliya ee AI oo leh moodelo la hawlgeliyay oo ku jira nidaamyada sirta ah ee Pentagonka. Kadib markii wada-hadalladii la falanqeeyay bishii hore ay fashilmeen, OpenAI, oo ah tartame weyn oo Anthropic ah, ayaa ku dhawaaqday heshiis Pentagon ah. Isla maalintii warqadda loo qaybiyay, Anthropic waxay dacwo kala soo dageen dowladda federaalka, iyagoo ku dooday in magacaabista khatarta silsiladda sahayda ay tahay aargudasho sharci darro ah oo lagula kacayo hadalka la ilaaliyo, iyagoo sheegay in aysan jirin wax sharci ah oo u oggolaanaya tallaabooyinka noocaas ah. Afhayeenka Aqalka Cad Liz Huston ayaa cambaareysay kiiska, iyadoo sheegtay in Madaxweyne Trump "waligiis oggolaan doonin shirkad cusub oo dhinaca bidixoodka ah iyo kuwa u heellan horumarinta siyaasadda in ay khatar geliso amniga qaranka by madax-bannaanida sida ay u shaqeyso ciidanka ugu xoogga badan uguna awooda badan aduunka. " Sida laga soo xigtay ilo yaqaan Claude oo ah mid muhiimad militari leh, waajibaadka ugu weyn ee Claude waa inay si degdeg ah u falanqeeyso warbixinnada sirdoonka iyada oo aqoonsanaysa qaababka, soo koobaysa natiijooyinka, iyo muujinaysa xogta muhiimka ah si ka waxtar badan dadka falanqeeya aadanaha. Rear Admiral Mark Montgomery oo hore u ahaa Taliyaha Ciidanka Badda, hadda ah aasaaska Foundation for Defense of Democracies, ayaa sheegay inay militerigu maalin kasta ka hawlgalaan ilaa kun oo bartilmaameedyo suurtagal ah waxayna badanaa ku dhacayso muddo afar saacadood gudahood, iyadoo AI ay awood u siineyso in heerkan iyo dhaqsahaas la gaadho iyadoo dadka ku lug leh go'aan qaadashada, taas oo aysan waligeed hore u dhicin.