- Amaphasiwedi asalokhu ebalulekile, kodwa i-AI yandise ikhono lokuwaqagela ngokuhlasela kwesichazamazwi, amandla amakhulu, kanye namamodeli okukhiqiza.
- Ama-LLM awakhiqizi ukungahleleki kwangempela: amaphethini awo abangela ukuthi amaphasiwedi awakhiqizayo abe ne-entropy ephansi futhi kube lula ngokumangazayo ukuwaphula.
- Ukuvikelwa kuhilela abaphathi bamaphasiwedi, amaphasiwedi amade nahlukile, ukuqinisekiswa kwezinto eziningi, kanye nokuqapha okuqhubekayo okusekelwe ku-AI.
- Ikusasa liqondiswe kumamodeli angenaphasiwedi, ama-passkey kanye ne-Zero Trust, lapho ubuwena kanye nomongo kubaluleke ngaphezu kwemfihlo eyodwa ehlanganyelwe.
Amaphasiwedi abhekene novivinyo lwangempela phakathi kwenguquko yobuhlakani bokwenziwa . Aseyisistimu yokuqinisekisa esakazeke kakhulu emhlabeni, kodwa yonke intuthuko ku-AI, ukuhlaselwa kwe-cyber, kanye nokwenza izinto ngokuzenzakalela kubeka ingcindezi eyengeziwe kumodeli ozalwe enkathini entsha. Okwamanje, izigidi zabasebenzisi ziyaqhubeka nokuvikela ama-akhawunti abo e-imeyili, amabhange, kanye nezinkundla zokuxhumana ngamaphasiwedi umhlaseli wanamuhla angawaphula ngemizuzwana.
Ngesikhathi esifanayo, akuwona wonke ama-akhawunti alungele ukungenela izwe elingenaphasiwedi . Ama-passkey, ukuqinisekiswa kwe-biometric , kanye namamodeli e-Zero Trust ayathuthuka, yebo, kodwa ahambisana namafomu okungena ngemvume akudala kanye nezinsizakalo zesikhathi esidlule ezisathembele kulokho "okuthile okwaziyo." Kulesi simo, ukuqonda ukuthi kungani amaphasiwedi ebaluleke kakhulu kunanini ngaphambili, ukuthi i-AI iwaphonsela kanjani inselelo, nokuthi yiziphi ezinye izindlela zangempela esinazo akuseyona inkinga yobuchwepheshe: kuyindaba yokusinda okuyisisekelo kwedijithali.
Kungani amaphasiwedi ehlala ebalulekile kule nkathi ye-AI
Namuhla, idatha yomuntu siqu neyenkampani isiphenduke imali eningi . Iphasiwedi eyodwa evuzayo ingathengiswa ngamadola ambalwa ezinkundleni ezingaphansi komhlaba, kodwa uma leyo phasiwedi ivula umnyango we-imeyili yenkampani, uhlelo lwamafayela lwangaphakathi, noma i-akhawunti yasebhange, umthelela wezezimali ungafinyelela amakhulu ezinkulungwane zama-euro.
Izinsizakalo ezahlukahlukene njengebhange eliku-inthanethi, i-imeyili, amarekhodi ezokwelapha, izinkundla zokuxhumana, kanye namathuluzi okusebenza asekelwe efwini asathembele kumaphasiwedi njengokulawula ukufinyelela okuyinhloko. Uma othile engena ngemvume nge-"password123" noma olunye uhlobo olungafanele, kufana nokushiya umnyango wakho wangaphambili uvuliwe ngophawu oluthi "ngena ungangqongqozanga".
Ucwaningo lubonisa ukuthi cishe ingxenye yazo zonke izinhlelo zokusebenza zithembele kuphela kumaphasiwedi , ngaphandle kwezinye izendlalelo zokuphepha. Futhi ngaphezu kwalokho, bambalwa abasebenzisi abasebenzisa amaphasiwedi ahlukene kusevisi ngayinye. Lokhu kuhluke kakhulu eqinisweni: abantu abaningi baphatha ama-akhawunti aku-inthanethi aphakathi kuka-70 no-100, kuyilapho thina, okungenani, singakhumbula amaphasiwedi ahlukene angu-5 kuya kwangu-7 kuphela.
Leli gebe phakathi kwalokho esingakukhumbula kanye nenani lama-akhawunti esiwasebenzisa nsuku zonke liholela emikhubeni eyingozi: ukusebenzisa kabusha amaphasiwedi, ukuwabhala phansi kumanothi angabethelwe, noma ukuwabelana ngeziteshi ezingaphephile . Nakuba kungenzeka ukuthi awukwenzi lokho, abantu abaningi basagcina amaphasiwedi kuhlelo lokusebenza lwamanothi noma ephepheni elinamathiselwe kusikrini sabo, okwenza kube lula kakhulu umsebenzi wanoma yimuphi umhlaseli ohlelekile kahle.
I-AI yengeza olunye ungqimba lobunzima: ukuhlaselwa kwe-brute-force , isichazamazwi, kanye nokufakwa kwemininingwane manje kungenziwa ngokuzenzakalelayo futhi kulungiselelwe ngezinga lezimboni . Lokho okwakudinga isikhathi, umzamo, kanye nobuchwepheshe obuthile manje kungahlelwa ngamathuluzi athuthukile ahlola izigidi zezinhlanganisela ezifundwe ekuphulweni kwedatha yangempela.
Ubude, ubunzima, kanye ne-entropy: yini eyenza iphasiwedi ibe namandla ngempela
Enye yezinganekwane ezisakazeke kakhulu ukuthi iphasiwedi "ebonakala iyinkimbinkimbi" ivikelekile ngokuzenzakalelayo . Ukwengeza izinhlamvu ezinkulu, ukufaka izinombolo esikhundleni sezinhlamvu, noma ukufaka uphawu kunikeza umbono othile wamandla, kodwa uma kuqhathaniswa nezindlela zokuhlasela zanamuhla, isici esinqumayo siyinto ehlukile: ubude.
Uma umhlaseli ethola ukufinyelela kusizindalwazi esivuvukile seziqinisekiso ezivela enkampanini enkulu, amaphasiwedi avame ukugcinwa hhayi embhalweni ocacile, kodwa njengama- hashes abethelwe . Ukuze bazame ukubuyisa iphasiwedi yokuqala, basebenzisa amasu okuqhekeka: bakhiqiza ukuqagela kwamaphasiwedi, bawabethele ngendlela efanayo, futhi baqhathanise imiphumela baze bathole okufanayo.
Kuleso simo, uma iphasiwedi iyinde, kulapho kukhula khona inani lezinhlanganisela ezingaba khona . Iphasiwedi enezinhlamvu ezingu-8 ingaqhekeka ngemizuzwana noma imizuzu ku-GPU cluster noma endaweni yokubala amafu, noma ngabe ishintsha izinhlamvu, izinombolo, kanye nezimpawu. Ngokuphambene nalokho, iphasiwedi enezinhlamvu ezingu-16 eyakhiwe kahle ingathatha iminyaka, amakhulu eminyaka, noma nje kube nzima ukuyiqhekeka ngezinsizakusebenza zamanje.
Uma sicabanga futhi ngamakhompyutha e-quantum esikhathi esizayo, ubude busasenza umehluko: umucu ongenangqondo njengohlamvu olufanayo oluphindwe ka-40 cishe akunakwenzeka ukuwuphula ngamakhono e-quantum enengqondo, kuyilapho iphasiwedi emfushane "yokudala" esekelwe eqenjini lakho lebhola noma engomeni edumile ingaqagelwa ngokushesha.
Ngokombono osemthethweni kakhulu, ukuphepha kwephasiwedi kulinganiswa ngama -bits e-entropy , okubonisa ukuthi zingaki imizamo engadingeka, ngokwesilinganiso, ukuze kuqagelwe. Uma isethi yezinhlamvu ezisetshenzisiwe ibanzi (izinhlamvu ezinkulu, izinhlamvu ezincane, izinombolo, izimpawu) futhi inhlanganisela ide, i-entropy iyanda futhi kuba nzima ukuqhekeka usebenzisa ukuhlaselwa kwe-brute-force.
Iphasiwedi enama-bits angaba ngu-20 e-entropy ikhiqiza cishe inhlanganisela engaba yisigidi, into umhlaseli onama-GPU anamuhla angayiqhekeza ngemizuzwana embalwa . Ngokuphambene nalokho, iphasiwedi enama-bits angaba ngu-100 e-entropy ingahlanganisa inhlanganisela eminingi kangangokuthi, empeleni, kungathatha izigidigidi zeminyaka ukuyisusa. Umehluko phakathi kwalokhu okubili awusiwo umlingo: ubude bangempela kanye nokuhlukahluka okusebenzayo kwezinhlamvu, hhayi nje ukubonakala kobunzima.
Iphutha elikhulu: ukushiya ukuphepha ezandleni zamaphasiwedi akhiqizwe yi-AI
Njengoba ama-chatbot esanda, kuyaheha ukucela imodeli efana ne-ChatGPT, i-Claude, noma i-Gemini ukuthi "ikhiqize iphasiwedi evikelekile enezinhlamvu ezingu-16 enezinombolo, izimpawu, nezinhlamvu ." Ephepheni, kuzwakala kuphelele: imodeli ibuyisa izintambo eziyinkimbinkimbi, okunzima ukuzifunda futhi okungenakwenzeka ukuzikhumbula ngokushesha, ezifana kakhulu nalezo ezinikezwa abaphathi bamaphasiwedi.
Inkinga iwukuthi, ngaphansi kwe-hood, amamodeli olimi awaklanyelwe ukukhiqiza ukungahleliwe kwangempela . Asebenza ngokubikezela umbhalo olandelayo (ithokheni) ngokusekelwe kudatha abaqeqeshwe kuyo. Lokhu kuphambene ngqo nalokho esikudingayo ukuze sibe nephasiwedi eqinile, edinga izinketho ezinokwenzeka ngokulinganayo nezingenakubikezelwa.
Ucwaningo lwakamuva lochwepheshe bezokuphepha be-AI lubonise ukuthi lapho la mamodeli ecelwa ukuthi akhiqize amaphasiwedi amaningi "ahlukile", kwavela amaphethini acacile kakhulu futhi aphindaphindayo . Ekuhlolweni kukaClaude, amaphasiwedi amaningi aqala ngohlamvu olufanayo, kanti uhlamvu lwesibili cishe njalo lwaluyinombolo efanayo; ngaphezu kwalokho, isethi encane kakhulu yamagama yasetshenziswa kabusha, kwasala izinhlamvu eziningi.
Kwenzeka into efanayo nge-ChatGPT: cishe wonke amaphasiwedi aqala ngohlamvu olufanayo, futhi cishe ingxenye yawo yasebenzisa u-"Q" njengomlingiswa wesibili . I-Gemini iphinde yabonisa amaphethini aphindaphindayo, anezinhlanganisela ezazibonakala zingahleliwe kodwa, lapho zihlaziywa ngokwezibalo, zazikude kakhulu nokusatshalaliswa okufanayo.
Into ephawuleka kakhulu ukuthi amaningi alawa maphasiwedi akhiqizwe awazange abonise ukuphindaphinda okukodwa kwezinhlamvu . Emehlweni omuntu, lokhu kunikeza umuzwa wokuhleleka okungahlelekile, kodwa ngokombono onokwenzeka, akunakwenzeka kakhulu ukuthi izinqumo bezithathwa ngokungahleliwe ngempela. Lokhu kuyinkomba ecacile yokuthi imodeli ilandela "imithetho yangaphakathi" eyenza ibikezelwe kakhulu kunalokho okubonakala.
Lapho i-entropy yala maphasiwedi adalwe yi-LLM ilinganiswa, imiphumela yayimbi kakhulu: uma kuqhathaniswa nama-bits angu-6,13 e-entropy ngohlamvu ngalunye oluzokhiqizwa yi-generator engahleliwe ngempela , amamodeli aphatha kuphela ama-bits angu-2,08 ngohlamvu ngalunye. Kuphasiwedi enezinhlamvu ezingu-16, lokho kusho ukusuka kuma-bits angaba ngu-98 (aqinile kakhulu) kuya kuma-bits angu-27 kuphela, into umhlaseli angayiphula ngokuhlasela okunamandla ngaphandle kokujuluka.
Ekugcineni, abacwaningi bathole amaphethini ahlukile kumaphasiwedi akhiqizwe yi-AI kuzo zonke izinsizakalo ze-inthanethi zangempela . Ngamanye amazwi, kunezinhlelo zokusebenza namawebhusayithi asevele evikelwe amaphasiwedi abelana ngezakhiwo ezibikezelwayo, okwenza kube yizinhloso eziphelele zokuhlaselwa kwezichazamazwi ezikhethekile ezifaka "uhlu lwamaphasiwedi ajwayelekile e-chatbot."
Yingakho izinkampani eziningi zokuphepha zicacile ngalokhu: akuwona umqondo omuhle ukudlulisela ukudalwa kwamaphasiwedi ku-AI ejwayelekile . Alungiselelwe ukukhiqiza umbhalo onengqondo, hhayi ukuqinisekisa ukusatshalaliswa okungahleliwe okumelana nokuhlaziywa. Uma usuvele ukhiqize amaphasiwedi nge-chatbot, kufanelekile ukuwashintsha futhi, okungenani, ukwengeza ukuqinisekiswa kwezinto ezimbili (2FA) ukuze kwandiswe izinga lokuvikelwa.
Abaphathi bamaphasiwedi, ukungahleliwe kwangempela, kanye nemikhuba emihle eyisisekelo
Uma ukukhumbula amaphasiwedi amaningi ahlukile, amade, futhi ayinkimbinkimbi kungenzeki (okuyinto engenakwenzeka), ikhambi elisebenzayo ukufaka ukuphathwa ndawonye . Yilapho abaphathi bamaphasiwedi bengena khona , bedala futhi begcina amaphasiwedi aqinile "endlini" ebethelwe evikelwe yiphasiwedi eyinhloko, futhi, ngokuya, yizici ezengeziwe zokuqinisekisa.
Ngokungafani nama-LLM, abaphathi bamaphasiwedi basebenzisa ama-cryptographic random number generator . Lawa athembele ku-operating system entropy kanye nemithombo ebonakalayo (ukunyakaza kwegundane, isikhathi phakathi kwemicimbi, njll.) ukukhiqiza ama-bits angalindelekile ngempela, abese eguqulwa abe yizinhlamvu ngaphandle kokwethula noma yimaphi amaphethini "amahle" esweni lomuntu.
Amathuluzi afana ne -Bitwarden, enikeza inguqulo yamahhala ephelele, noma i-1Password, isevisi ekhokhelwayo kodwa enamandla kakhulu , akuvumela ukuthi uphathe hhayi amaphasiwedi kuphela, kodwa namakhadi esikweletu, amanothi aphephile, nolunye ulwazi olubucayi. Umsebenzisi udinga kuphela ukukhumbula iphasiwedi enamandla kakhulu (futhi, ngokufanele, ukuvikela i-akhawunti ngokuqinisekiswa kwezinto ezimbili), kuyilapho umphathi enakekela okunye.
Ngaphezu kokudala amaphasiwedi aqinile, laba baphathi bavame ukufaka izici ezifana nokujikeleza kweziqinisekiso njalo, izexwayiso zokwephulwa kwedatha okwaziwayo, kanye nokuhlolwa kokusetshenziswa kabusha kwamaphasiwedi . Uma iwebhusayithi iphulwa noma uma idatha yakho ivela ohlwini lwewebhu olumnyama, umphathi angakuxwayisa ukuze ukwazi ukushintsha amaphasiwedi athintekile ngaphambi kokuba othile awasebenzise kabi.
Omunye umkhuba omuhle ukususa ama-akhawunti ongasawasebenzisi . Kulula ukukhohlwa ngephrofayela endala ye-Myspace, i-akhawunti ye-Hotmail yasekuqaleni kwawo-2000, noma isevisi oyibhalisele "ukuze nje uyizame." Kodwa-ke, uma wake wasebenzisa lelo kheli le-imeyili njengekheli lokubuyisa elinye i-akhawunti ebalulekile, liba yisixhumanisi esisengozini kuchungechunge lwakho lokuphepha.
Okokugcina, ukuvumela ukuqinisekiswa kwezinto eziningi kuwo wonke ama-akhawunti angaba khona kuyimpoqo . Kungakhathaliseki ukuthi usebenzisa i-SMS, izinhlelo zokusebenza ezifana ne-Authy noma i-Google Authenticator, noma izindlela ezithuthukisiwe kakhulu, ezidinga isici sesibili (ikhodi, ukhiye ongokoqobo, i-biometrics) kunciphisa kakhulu umthelela wephasiwedi eyonakalisiwe.
Indlela i-AI enika ngayo amandla abahlaseli kanye nabavikeli
Ubuhlakani bokwenziwa bushintsha umdlalo kuzo zombili izinhlangothi. Ngakolunye uhlangothi, izigebengu ze-inthanethi azisathembeli kuphela emandleni akudala : ziqeqesha amamodeli anezigidi zamaphasiwedi avuzayo ukuze akhiqize izichazamazwi ezisebenza kahle kakhulu, ezikwazi ukubamba amaphethini abantu lapho kwakhiwa amaphasiwedi (amagama, izinsuku, amaqembu, amagama ezingoma, njll.).
Amathuluzi afana ne-PassGAN, asekelwe kumanethiwekhi akhiqizayo, angakhiqiza inani elikhulu lokuqagela iphasiwedi ngaphandle kokudinga imithetho ecacile, echazwe ngabantu . Ngokungafani nezinsiza zendabuko ezifana ne-Hashcat, ezisebenzisa izinguquko ezichazwe kusengaphambili ohlwini olwaziwayo, lawa mamodeli ayaqhubeka nokufunda nokuthuthukisa imiphumela yawo njengoba enza imizamo eminingi.
Ekuhlolweni okwenziwe ngamasethi edatha afana namaphasiwedi e -RockYou leak noma e-LinkedIn adumile, i-PassGAN iqagele kahle phakathi kwamaphasiwedi ahlukile angu-51% no-73% kunamathuluzi asekelwe emithethweni . Iqeqeshwe ngokukhethekile kumasampula avela ekuvuzeni okuthile, ikwazile ukuqagela kahle cishe ingxenye yesine yamaphasiwedi kwenye isethi ezimele—isibalo esikhathazayo kakhulu uma sicabangela abasebenzisi abasebenzisa kabusha amaphasiwedi kuzo zonke izinsizakalo ezahlukene.
Ngaphezu kwalokho, amasu okuhlasela njengokufaka imininingwane kanye nokufutha amaphasiwedi kusizakala kakhulu ngokuzenzakalelayo okuthuthukisiwe . Nge-AI, abahlaseli bangalungisa ukuhleleka kanye nokubaluleka kwamaphasiwedi abawazamayo, bathuthukise izinsiza futhi bagxile ekuhlanganisweni okunamathuba aphezulu okuphumelela ngokusekelwe kuphrofayela yesisulu.
Kodwa i-AI iphinde iphindaphinde umonakalo ngenye indlela: i-phishing kanye nobunjiniyela bezenhlalo . Amamodeli akhiqizayo avumela ukudalwa kwama-imeyili, amawebhusayithi okungena ngemvume mbumbulu, ngisho nezingcingo zezwi ezinama-deepfake alingisa amalungu omndeni, abaphathi, noma ozakwabo abanezinga eliphazamisayo lokungokoqobo. Kulezi zimo, kungakhathaliseki ukuthi iphasiwedi yakho iqine kangakanani, ayisizi ngalutho uma ugcina uyithayipha ekhasini mbumbulu eliyithumela ngqo kumhlaseli.
Nokho, ohlangothini lokuzivikela, i-AI isibe umlingani obalulekile . Izinhlelo eziningi zokuphepha zisebenzisa amamodeli ukuhlaziya amaphethini okungena ngemvume nokuthola ukuziphatha okungavamile: ukufinyelela okuvela emazweni angajwayelekile, izikhathi ezisolisayo, isivinini sokuthayipha esingahambisani nomuntu, noma amadivayisi angajwayelekile.
Uma, isibonelo, ukufinyelela kutholwa endaweni engajwayelekile kuhlanganiswe nephethini yokusebenzisana ebonakala sengathi izenzakalelayo , uhlelo lungavimba umzamo, lucele ukuqinisekiswa okwengeziwe, noma luphoqe ukushintsha iphasiwedi. Isihluthulelo lapha ukuqapha okuqhubekayo: ukukhiya nje umnyango akwanele; kufanele uqaphele ukuthi ubani ozama ukungena nokuthi kanjani.
Ngale kwephasiwedi: i-MFA, i-biometrics, i-Zero Trust kanye nama-passkey
Njengoba ukwanda okuqhubekayo kokuhlaselwa kanye nokwanda kobunikazi bedijithali (kokubili obungumuntu nobomuntu), imodeli yakudala esekelwe kuphela kumagama abasebenzisi namaphasiwedi iye yahluleka . Umkhuba uwukushintsha umthwalo wokuphepha kusuka emfihlweni eyodwa uye ekuhlanganisweni kwezimpawu, umongo, kanye nokuqinisekiswa okuqhubekayo.
Okokuqala, ukuqinisekiswa kwezinto eziningi (i-MFA) sekuyinto ebalulekile . Akuzona zonke izici eziqinile ngokulinganayo: amakhodi esikhathi esisodwa nge-SMS, isibonelo, asengozini yokuhlaselwa yi-interception, ukwenziwa kwe-SIM card cloning, noma ubugebengu bokweba imininingwane ebucayi. Noma kunjalo, angcono kunokungenzi lutho, futhi isinyathelo esilandelayo ukutshala imali ezindleleni eziqinile, njengezinhlelo zokusebenza ezizinikele, okhiye bokuphepha, noma izaziso zokucindezela ezivikelwe ukukhwabanisa.
I-biometrics, njengokubona ubuso noma ukuskena izigxivizo zeminwe kumafoni anamuhla nama-laptop , inikeza ulwazi olulula kakhulu futhi luphakamisa izinga labahlaseli, yize lungavimbeli. Uma ihlanganiswe nezinkinobho ze-cryptographic ezigcinwe kudivayisi, ivumela izinhlelo lapho kungekho khona iphasiwedi yomsebenzisi yokuthayipha, okunciphisa ingozi yobugebengu bokweba imininingwane ebucayi.
Yilapho-ke ama-passkey nezixazululo ezingenaphasiwedi zisebenza khona , kusetshenziswa i-cryptography yomphakathi ukuqinisekisa umsebenzisi ngaphandle kwesidingo sokhiye wendabuko. Izinsizakalo ezifana ne-Okta FastPass zikuvumela ukuthi uxhume idivayisi ngokuphephile kubunikazi futhi uqinisekise ukufinyelela usebenzisa okhiye be-cryptographic, futhi zihlanganiswe nezindlela ezifana ne-Windows Hello, i-Apple Touch ID, noma i-Face ID.
Lolu hlobo lwesixazululo alugcini nje ngokuthuthukisa ulwazi lomsebenzisi (akusekho ukukhumbula amaphasiwedi amaningi), kodwa futhi lunciphisa kakhulu amathuba okuthi umhlaseli antshontshe iziqinisekiso ezingasetshenziswa kabusha . Akukho "imfihlo eyabiwe" ehamba kunethiwekhi engabanjwa: ukuqinisekiswa kwenziwa ngezinselele ze-cryptographic ezingaxazululwa yidivayisi esemthethweni kuphela.
Ngesikhathi esifanayo, izinhlangano eziningi zamukela indlela ye-Zero Trust, ecabanga ukuthi akukho ukufinyelela okuthembekile ngokuzenzakalelayo , noma ngabe kuvela ngaphakathi kwenethiwekhi yenkampani. Isicelo ngasinye sihlolwa ngokubheka izici eziningi: ukuthi ungubani umsebenzisi, iyiphi idivayisi ayisebenzisayo, indawo yakhe, isikhathi sokufinyelela, ukuziphatha kwakhe kwakamuva, njalo njalo.
Kulo modeli, ukuphathwa kobunikazi nokufinyelela (i-IAM/IGA) kuba yinto ebalulekile ekulawulweni kokuphepha . Izimiso zamalungelo aphansi, ukubuyekezwa kwemvume ngezikhathi ezithile, kanye nokuxhumana kwesikhathi sangempela kwezimpawu zengozi kuyasetshenziswa. Uma okuthile kuphambuka kuphethini elindelekile, uhlelo lungahoxisa amalungelo ngokushesha noma ludinge ukuqinisekiswa okwengeziwe.
Icala elisha: Ubunikazi be-ejenti ye-AI kanye nama-akhawunti angewona awabantu
Ukuvela kwe-AI ngama-ejenti azimele kunezela olunye ungqimba lobunzima. Lawa ma-ejenti awagcini nje ngokubuza ngezinhlelo, kodwa futhi asebenza ngaphakathi kwawo : ayaqinisekisa, abize ama-API, afunde futhi abhale idatha, futhi aze enze nezinqumo zokusebenza egameni lenkampani.
I-ejenti ngayinye yalolu hlobo idinga ubunikazi bayo noma, okungenani, isethi yeziqinisekiso . Lokhu kwandisa kakhulu indawo yokuhlasela, ngoba sikhuluma ngokhiye be-API, amathokheni e-OAuth, ama-akhawunti esevisi, nezinye izimfihlo ezivame ukungafakwa ohlwini olubi, ezinezimvume eziningi, noma eziphelelwa yisikhathi sekwephuzile (uma zikhona).
Ngaphezu kwalokho, izinhlelo ezisekelwe kuma-ejenti zethula ubuthakathaka obusha nobuthile : ukufaka umyalo emiyalweni abayitholayo, ukufaka ubuthi kumamodeli, ukushintshwa kwedatha, ukwenyuswa kwamalungelo angadingeki, njalo njalo. Akukhona nje kuphela ukuthi ikhodi yendabuko iyahlaselwa, kodwa futhi ne-logic yokwenza izinqumo ye-AI.
Lokho okubizwa ngokuthi “i-shadow AI” kwenza isimo sibe sibi kakhulu: abasebenzi abahlanganisa amathuluzi e-AI angagunyaziwe emisebenzini yabo yokusebenza bangadala ubuthakathaka obungadalelwanga ukubhekana nokulawula kwendabuko (izicishamlilo, i-antivirus, izinqubomgomo zefa). I-ejenti engalungiselelwe kahle, isibonelo, ingahoxisa iziqinisekiso ezisemthethweni noma ihlukanise izinsiza ezibalulekile ngaphandle kokuthi othile aqaphele ngesikhathi.
Ngakho-ke, ochwepheshe batusa ukuphatha ama-ejenti e-AI njengobunikazi bedijithali obubiza kakhulu . Lokhu kuhilela ukubanikeza izimvume ezidingekayo kuphela, ukuqapha eduze umsebenzi wabo, ukusungula izindlela zokumisa noma zokunciphisa izenzo zabo uma bephambuka ekuziphatheni okulindelekile, nokuqinisekisa ukuthi ukusekela ngokulondoloza, ukululama, kanye nokuphatha nakho kubhekana nalezi zingxenye.
Inselele enkulu iwukubonakala okusobala: izinkampani eziningi azazi nokuthi zingaki ama-ejenti asebenza nganoma yisiphi isikhathi, ukuthi axhuma kuziphi izinhlelo, noma ukuthi afinyelela kuluphi ulwazi . Ngaphandle kwempahla noma ukulandelelwa, ingozi yokulahlekelwa ukulawula iyanda kakhulu, nanoma yisiphi isigameko singandiswa ukuzimela kwalezi zinhlelo.
Izindaba ezinhle ukuthi amathuluzi okuphatha ubunikazi, izimvume, kanye nokufinyelela—kokubili okungabantu kanye nokungebona abantu—asevele ekhona futhi asevuthiwe . Inselele yangempela ingokwesiko nokuhlela: ukuhlanganisa ukubusa, ukuqina, kanye nokuqeqeshwa kusukela ekuqaleni, kunokuphatha i-AI njenge-plugin elula ye-plug-and-play.
Uma sibheka isithombe esikhulu, kusobala ukuthi amaphasiwedi asadlala indima ebalulekile, kodwa awasakwazi ukuqinisekisa ukuphepha kwedijithali ngawodwa . Esimweni lapho i-AI isheshisa khona kokubili ukuhlasela nokuzivikela, lapho ama-ejenti azimele enza izinqumo ezinhlelweni ezibucayi, nalapho ubunikazi buphindaphindeka khona, isu lihilela ukusebenzisa amaphasiwedi amade, ahlukile aphethwe ngamathuluzi aqinile, ewasekela njalo nge-MFA, enciphisa kancane kancane ukubaluleka kwawo ngama-passkey kanye nezinhlelo ezingenaphasiwedi, kanye nokuqinisa ubunikazi kanye nemodeli yokufinyelela esekelwe ku-Zero Trust kanye nokuqapha okuqhubekayo. Kungale ndlela kuphela lapho singagcina khona indawo esongelayo eba ngokushesha, ezenzakalelayo kakhulu, futhi ehlakaniphe kakhulu nsuku zonke.
