Ukwesaba ukuhlasela kwama-cyberattack kubambezela i-AI, kodwa izinkundla ezifana ne-Google Gemini zinikeza indlela ephephile, ochwepheshe kwa-L8 Group bayasho.

Ngokocwaningo lwe-TOTVS, ama-36% ezinkampani zase-Brazil zisangabaza ukutshala imali ku-Artificial Intelligence ngenxa yokwesaba ukuhlaselwa kwe-inthanethi nokuphulwa kwedatha. Kodwa-ke, ukuthuthukiswa kwezinkundla ezivikelekile ngokwemvelo, njengeGemini ye-Google, esingethwe kwenye yengqalasizinda yamafu enamandla kakhulu emhlabeni, kushintsha le paradigm. Kochwepheshe bakwa-L8 Group, inkampani ehamba phambili yobuchwepheshe kanye ne-cybersecurity, ukhiye wokuvikela ukusungula izinto ezintsha usekwazini ukuthi asetshenziswa kanjani amandla okuvikela lezi zinkundla esezivele ziwanikeza.

"Ukwesaba ukuveza idatha ebucayi kanye nokuvula iminyango emisha ezinsongweni kuyisithiyo esikhulu ekuthathweni kwe-AI; kodwa-ke, ukukhetha kwenkundla yobuchwepheshe kuyisici esinqumayo ekunciphiseni lezi zingozi. Ukwesaba kwezinkampani kuyaqondakala, kodwa kubangelwa umbono wokuthi i-AI iyibhokisi elimnyama elisengozini. Lokhu akulona iqiniso uma sikhuluma ngamamodeli afana ne-Gemini. Akuyona i-ecoclasse-class esebenza vele ohlelweni lwezokuphepha lwe-Google; izingqimba," kuchaza u-Guilherme Franco, i-CTO ye-L8.

Lokhu kusho ukuthi idatha yekhasimende ivikelwe ukubethela okuthuthukile, izinqubomgomo zobumfihlo eziqinile ezivimbela ukusetshenziswa kwayo ekuqeqesheni amamodeli asesidlangalaleni, kanye neqoqo lamathuluzi okulawula. NgokukaFranco, ezokuphepha akusona isengezo; isisekelo, esingenziwa ngokwezifiso nakakhulu uma izinkampani sezivele zisebenzisa i-Google Workspace, ziyihlanganisa nezinqubomgomo ze-Vault zokugcinwa kwedatha, isibonelo.

Ezinkampanini ezifuna ukutshala imali ku-AI ngokuphephile zisebenzisa i-Gemini, i-L8 Group igcizelela ukuthi impumelelo incike ekucushweni okufanele nasekukhuliseni izici zokuphepha ezitholakala kuplathifomu ye-Google Cloud. Nawa amanye amaphuzu aphakanyiswe uchwepheshe we-cybersecurity u-Guilherme Franco:

1. Vikela Ingqalasizinda Ngokuzenzakalelayo: I-Gemini izuza kungqalasizinda efanayo evikela i-Gmail, Usesho, ne-YouTube. Lokhu kubandakanya ukuvikelwa kwe-denial-of-service (DDoS), ukutholwa kokungena, kanye nenethiwekhi yomhlaba wonke eyimfihlo, ebethelwe.
2. Ukulawulwa Kwedatha Nokufinyelela (i-IAM ne-VPC-SC): I-Google Cloud Identity and Access Management (IAM) ikuvumela ukuthi ucacise ukuthi ubani ongafinyelela amamodeli nedatha ye-AI. Ukwengeza, Nge-VPC Service Controls, izinkampani zingakha umjikelezo wokuvikeleka obonakalayo ukuvimbela ukuvuza kwedatha, ziqinisekise ukuthi ulwazi olubucayi alushiyi indawo elawulwayo.
   1. Kubasebenzisi be-Google Workspace, i-Gemini ihlonipha amaleveli okufinyelela afanayo ayechazwe ngaphambilini ukuze afinyelele okuqukethwe kwenkampani, okufana ne-Google Drayivu, ngaphandle kwesidingo sokucushwa okwengeziwe.
   2. Okufanayo kunganwetshwa nakubasebenzisi abasebenzisa izinkundla ngaphandle kwe-Google Workspace, njenge-Microsoft, uma usebenzisa i-Google Agentspaces nge-IAM ethuthukisiwe.
3. Ubumfihlo Nokugcinwa Kuyimfihlo: I-Google iqinisekisa ngokwenkontileka ukuthi idatha yebhizinisi efakwe ku-Gemini nge-Google Cloud ngeke isetshenziselwe ukuqeqesha amamodeli afinyeleleka esidlangalaleni. Ukulawula nobunikazi bedatha kuhlala ngokuphelele enkampanini yeklayenti.
4. Ukuphepha Kwe-AI Okunesibopho Nezihlungi: Inkundla ye-Gemini ngokwayo inezihlungi zokuphepha ezakhelwe ngaphakathi ukunciphisa ukukhiqizwa kokuqukethwe okungalungile, okuyingozi, noma okuchemayo, okungavikeli idatha kuphela kodwa nesithunzi somkhiqizo.
5. Idatha "Yasendaweni": Kungenzeka ukusebenzisa amathuluzi afana ne-NotebookLM, phakathi kokunye, abheka okuqukethwe ngokufunda amafayela akhethwa ngumsebenzisi, ngaphandle kokusebenzisa isizindalwazi sokusesha sangaphandle, njenge-inthanethi, ukunciphisa ukukhohliswa kanye nokuqinisekisa ubumfihlo obukhulu.

Okokugcina, uchwepheshe uyaxwayisa: "Umbuzo awusekho 'uma' sizosebenzisa i-AI, kodwa 'kanjani' sizokwenza ngokuphepha nangengozi. Amapulatifomu afana ne-Gemini axazulula inkinga enkulu yokuphepha emnyombweni. Umsebenzi wethu ku-L8, isibonelo, ukwenza njengozakwethu wamasu owenza ngokwezifiso futhi asebenzise lezi zingqimba zokuvikela: IAM, i-VPC idinga ukuguqulwa kweqiniso lebhizinisi ngalinye. amandla e-AI abe yinzuzo yokuncintisana evikelekile futhi elungele ikusasa Okubaluleke kakhulu, sakha amaphrojekthi asebenza ngempela, ngoba ucwaningo lwakamuva lwe-MIT luveze ukuthi amaphrojekthi we-AI angama-95% ayahluleka, "kwengeza uFranco.

Uphinde axwayise ngokuthi, mayelana ne-cybersecurity, ngaphezu kwegama eselivele lalaziwa elithi ShadowIT, kukhona ne-ShadowAI, lapho abasebenzisi basebenzisa khona amathuluzi e-AI angagunyaziwe futhi angavikelekile. "Ezinye izinkundla ziqeqesha ama-AI azo ngokusekelwe ekufakweni komsebenzisi, okuhlanganisa idatha ebucayi, ephula i-LGPD. Cabangela icala lakamuva le-Grok, elivuze izingxoxo eziyimfihlo ezingaphezu kuka-370,000. Ukusiza ukuthola nokuvimbela ukusetshenziswa kwe-ShadowIT ne-ShadowAI, i-L8 Group inikeza izixazululo ezihlinzeka ngokubonakala nokulawula lokho okufinyelelwayo, ngokuvumelana nezinqubomgomo ze-cybersecurity.