Anthropic ta sanar da cewa za ta "katse aikin" na na'urorin AI na ci gaba ga dukkan masu amfani bayan da gwamnatin Amurka ta umarci kamfanin da ya hana 'yan kasashen waje shiga, saboda dalilan tsaron kasa.
Kamfanin ya sami umarnin sarrafa fitarwa don dakatar da shiga Fable 5 da Mythos 5 ga dukkan 'yan kasashen waje, amma ba a ba shi cikakkun bayanai game da matsalar tsaron kasa ba, in ji Anthropic a cikin wata sanarwa.
A cewar kamfanin, ya fahimci cewa gwamnati ta yi imanin akwai hanyar da za a bi, ko "jailbreak," wani tsari da aka tsara don hana Fable 5 yin amfani da shi don gano raunin software.
Wannan umarni ya zo ne a daidai lokacin da wani rikici na baya tsakanin jami'an gwamnatin Trump da Anthropic—wanda ke shirin yin IPO—ya bayyana yana raguwa a wasu sassan gwamnatin Amurka.
Dangantakar Anthropic da gwamnati ta lalace a farkon wannan shekara bayan da ta ki yarda da sojojin Amurka su yi amfani da na'urorin AI don sa ido na cikin gida da kuma tsarin makamai masu sarrafa kansu. A matsayin martani, gwamnati ta sanya Anthropic a cikin jerin baƙar fata na sarkar samar da kayayyaki, wanda zai fara aiki daga baya a wannan shekara.
Matakin kuma ya nuna babban haɓaka a ƙoƙarin Amurka na iyakance iyawar AI na abokan gaba na kasashen waje. Shekaru da yawa, sarrafa fitarwa na Amurka ya mayar da hankali kan kwakwalwan kwamfuta da kayan aikin da ke ba da AI, maimakon hana shiga AI kanta ga kasashen waje.
Anthropic ta ce gwamnati ta ba da "shaida ta baki kawai game da yiwuwar jailbreak mai iyaka, wanda ba na duniya ba."
"Mun ƙi yarda cewa gano wani jailbreak mai iyaka ya kamata ya zama dalilin tuno da wani samfurin kasuwanci da aka tura ga ɗaruruwan miliyoyin mutane," in ji kamfanin.
Umarnin gwamnati da martanin Anthropic sun nuna karuwar tashin hankali tsakanin masu haɓaka AI da masu kula da su game da yadda za a tantance haɗarin daga abin da ake kira "jailbreaks"—hanyoyin da ake amfani da su don ketare tsare-tsaren kariya na samfurin.
Har zuwa ranar Laraba, Anthropic ta yi kira ga gwamnatin Amurka da ta kara sa ido kan AI, gami da ikon toshe samfuran da ke da haɗarin da ba a yarda da su ba. Duk da haka, ta ce matakin gwamnati a ranar Juma'a bai bi ka'idodin gaskiya da tsari na gaskiya ba.
Babban jami'in bayanai na Pentagon, Kirsten Davies, ta ce a wani rubutu a kan X cewa ma'aikatar tsaro ta goyi bayan fifikon tsaron kasa.
"Wasu abubuwa sun fi muhimmanci fiye da zagayowar kudaden shiga, clickbait da kimar kafin IPO. Amurka ta Farko. Koyaushe," in ji Davies.
Anthropic ta shigar da takardar IPO ta Amurka a asirce a watan jiya, ta ci gaba da gaban abokin hamayyarta OpenAI a tseren zuwa kasuwannin jama'a.
A farkon wannan mako, Anthropic ta ƙaddamar da wani samfurin AI mai suna Claude Fable 5, wanda ke wakiltar sabon matakin iyawa da take kira "Mythos-class." Samfurin ya haɗa da shinge da ke hana amfani da shi a wurare masu haɗari kamar tsaron yanar gizo, wanda wasu masu amfani suka soki a matsayin "mai yawa," in ji Anthropic.
Masana sun yi gargadin cewa samfuran Mythos, idan aka yi amfani da su ba daidai ba, na iya haɓaka hare-haren yanar gizo masu rikitarwa sosai, musamman a sassa kamar banki waɗanda ke dogaro da tsarin fasaha masu rikitarwa, masu haɗin gwiwa, kuma galibi tsofaffi.
Anthropic ta ce ta yi aiki tare da gwamnatin Amurka da sauran su kan aminci kafin ƙaddamar da Fable, kuma cewa samfuran daga masu samar da AI masu hamayya sun nuna irin wannan iyawar gano ƙananan kurakurai a cikin lambar.
"Sakamakon wannan umarni shine cewa dole ne mu katse aikin Fable 5 da Mythos 5 ga dukkan abokan cinikinmu don tabbatar da bin doka. Shiga duk sauran samfuran Anthropic ba zai shafa ba," in ji Anthropic.
Anthropic ta kara da cewa ta yi imanin akwai "rashin fahimta" kuma tana aiki don maido da shiga samfuran da wuri-wuri.
Wani jami'in Amurka ya tabbatar da cewa ma'aikatar kasuwanci ta ba da umarnin sarrafa fitarwa don dakatar da duk shiga Fable 5 da Mythos 5 ta 'yan kasashen waje.
Tambayoyi da Amsoshi
Ga jerin tambayoyi da amsoshi game da yanayin hasashe inda Anthropic ta katse aikin na'urorin AI na ci gaba bayan umarnin gwamnatin Amurka da ke hana shiga kasashen waje
Tambayoyi na Gabaɗaya
Q Me yasa Anthropic ke katse aikin na'urorin AI na ci gaba
A Saboda sabon umarnin gwamnatin Amurka ya hana ƙungiyoyin kasashen waje shiga mafi ƙarfin tsarin AI Anthropic tana bin wannan umarni don bin doka
Q Wannan yana nufin ba zan iya amfani da Claude kwata-kwata ba
A A'a Wannan yana iya shafar samfuran mafi ci gaba kawai Wataƙila za ku iya amfani da nau'ikan Claude na yau da kullun ko marasa ƙarfi
Q Wannan na dindindin ne
A Ya dogara da umarnin Zai iya zama dakatarwa na ɗan lokaci yayin da gwamnati ke nazarin haɗarin tsaro, ko kuma yana iya zama hani na dindindin har sai dokokin sun canza
Q Wanene wannan hani ya shafa
A Da farko masu amfani, kamfanoni, ko masu haɓaka da ke zaune a wajen Amurka ko kuma mallakar ƙungiyoyin kasashen waje ne Masu amfani da ke Amurka ba su da tasiri
Tasiri akan Masu Amfani da Masu Haɓaka
Q Ina zaune a Kanada/Turai/Asiya Shin biyan kuɗin Claude na zai daina aiki
A Wataƙila za ku rasa shiga mafi girman matakin samfurin Ana iya rage asusunku zuwa samfurin mara ƙarfi, ko kuma za ku iya amfani da API kawai tare da tsofaffi/ƙananan samfuran
Q Ni mai haɓaka ne da ke amfani da API Shin ina buƙatar canza lambar ta
A E, wataƙila Kuna buƙatar sabunta kiran API don amfani da wani sunan samfurin daban Duba takardun Anthropic don takamaiman samfuran da aka yarda
Q Idan na yi amfani da VPN don nuna cewa ina Amurka fa
A Hakan zai keta sharuɗɗan sabis da sabuwar doka Anthropic za ta toshe asusun da aka gano suna amfani da VPN don ketare hani Kuna iya rasa duk asusunku
Q Shin za a share tattaunawa ko bayanan da nake da su
A Ba zai yiwu ba Tarihin asusunku da bayananku ya kamata su kasance, amma ba za ku iya samar da sabbin amsoshi ta amfani da samfurin da aka hana ba
Tambayoyi na Fasaha da Tsaro
