Ubuhlakani bokwenziwa budlulele ngale kokuba yisithembiso futhi sebuyimbangela yokuphepha kusiphequluli esithandwa kakhulu emhlabeni. I-Google imemezele ukuthi, ngenyanga kaJuni, i-Chrome ilungise ubuthakathaka bokuphepha obuyi-1.072 ezinguqulweni ezingu-149 no-150 , inani elidlula inani eliphelele lama-patches asetshenziswe ezinguqulweni ezingu-23 zangaphambilini zihlangene. Lesi sinyathelo esibalulekile asiyona into eyenzekile: inkampani ihlobanisa lokhu nohlelo lwe-AI agent olusekelwe ku-Gemini olushintshe indlela ubuthakathaka obutholakala ngayo, obuhlukaniswa ngayo, futhi obulungiswa ngayo.
Ithimba lezokuphepha le-Chrome belilokhu lisebenzisa ukufunda komshini iminyaka eminingi, kodwa ukufika kwamamodeli ezilimi amakhulu (i-LLM) kuvumele ukuthi bakwazi ukwenza ngokuzenzakalelayo izinqubo ezazidinga amahora amaningi omsebenzi wezandla . Ngokusho kwedatha ye-Google uqobo, ngoMeyi kuphela bavimbele ubuthakathaka obungaphezu kuka-20 ekufinyeleleni ekukhiqizeni, okuhlanganisa nephutha elibalulekile elase lisetshenziswa kakade . Lezi zindaba zibikwe yimithombo yezindaba eminingi, evumelana ngokuthi i-AI ichaza kabusha ukuphepha kwe-cyber ezingeni lezimboni.
Irekhodi elingakaze libonwe ekulungisweni
Izinombolo ziyazikhulumela. Izinguqulo ze-Chrome 149 kanye ne-150, ezikhishwe ngoJuni, zifaka phakathi ama-patches okuphepha angu-1.072, kanti izinguqulo zangaphambilini ezingu-23 zihlangene zaba ngu-1.036. Lokhu kusho ukuthi ubuthakathaka obuningi bulungisiwe ngenyanga eyodwa kunaseminyakeni emibili edlule . Umbiko wobuchwepheshe oshicilelwe yi-Google uchaza ukuthi abaphathi be-Large-Scale Vulnerability (LLMs) bavula amakhono angakaze abonwe okuthola ubuthakathaka okuzenzakalelayo, becindezela imingcele yobuchwepheshe babantu. Inkampani igcizelela ukuthi le ndlela ayithathi indawo yezindlela zendabuko, kodwa kunalokho iyazanelisa, ivumela ukumbozwa kwekhodi yomthombo ebanzi kakhulu.
Uhlelo Lwemiklomelo Yokuhlushwa (i-VRP) nalo luzwile umthelela. NgoMashi 2026, i-Google yayisivele ithole imibiko eminingi kunangesikhathi sonke sika-2025, okwabaphoqa ukuthi bashintshe imigomo yohlelo ukuze babeke phambili okutholakele okunikeza okuthile okuhlukile kulokho amathuluzi azenzakalelayo asevele ekuthola. I-AI ihlunga ugaxekile, iphinda ubufakazi bomqondo, futhi inikeze amazinga obunzima , isindise onjiniyela amahora amaningi omsebenzi inyanga ngayinye.
I-ejenti ye-AI ethole iphutha elineminyaka engu-13 ubudala
Enye yezinto eziphawuleka kakhulu ezitholwe ohlelweni olusha kwaba ukutholakala kwe-sandbox escape eyayifihliwe kukhodi ye-Chrome iminyaka engaphezu kwengu-13. Njengoba ibekwe kuhlu njenge-CVE-2026-3545, ine-CVSS score engu-9.8 (ebucayi), leli phutha lalizovumela i-renderer esengozini ukuthi ikhohlise isiphequluli ukuthi sifunde amafayela endawo kusuka kukhompyutha. I-ejenti esekelwe ku-Gemini, eyadalwa ekuqaleni kuka-2026, yayinesibopho sokuyithola ngokuhlaziya ikhodi yomthombo ephelele ye-Chrome ezingeni lemininingwane amaqembu abantu angenakuyigcina ngendlela eqhubekayo.
I-Google ichaza ukuthi uhlelo luhlanganisa ama-ejenti amaningi akhethekile: amanye askena i-codebase efuna amaphethini e-semantic, amanye aqinisekisa ukuthi ubuthakathaka bungokoqobo, kanti amanye akhiqiza ama-patches afanelekayo. Ukubuyekezwa kwabantu kusalokhu kuyimpoqo ngaphambi kokusebenzisa noma yiziphi izinguquko , kodwa i-AI isheshisa kakhulu inqubo. UDoug Turner, umqondisi wobunjiniyela kwa-Chrome, uthe ama-LLM ashintshe ngokuyisisekelo ezomnotho zokuphepha kwe-cyber, aguqula ukutholakala kobuthakathaka kube ukusebenza okuzenzakalelayo, kwezimboni.
Indlela uhlelo lwe-ejenti ye-Gemini olusebenza ngayo
Uhlelo lwe-Google aluyona imodeli eyodwa, kodwa luwukuhlelwa kwama-ejenti asebenza ochungechungeni. Ama-ejenti okuthola askena ikhodi ye-Chrome ukuthola amaphethini angase abonise ubuthakathaka, esebenzisa ukuqonda ikhodi ye-semantic, hhayi amaphethini ombhalo kuphela. Ama-ejenti okuqinisekisa abese eqinisekisa ukuthi okutholakele kuyiqiniso noma kuyiphutha. Uma sekuqinisekisiwe, ama-ejenti okulungisa akhiqiza ama-patches amaningi angaba khona , kuyilapho ama-ejenti okuhlola ama-patch ewaqhathanisa nemithetho nemigomo yesitayela sephrojekthi ye-Chromium.
Lonke lolu hlelo lusebenza ngekhodi egcinwe emishinini ekhiyiwe, ngaphandle kokufinyelela i-inthanethi okuvamile, ukuze kuncishiswe izingozi. I-Google iphinde yahlanganisa amathuluzi e-AI ohlelweni lwayo lokuhlanganiswa okuqhubekayo (CI), olusebenza njalo emahoreni angama-24 kuwo wonke amazinga okwakha. Lokhu kuvumela ukutholakala okusheshayo kwamaphutha okuphepha ngaphambi kokuba afinyelele ekukhiqizweni . Le nkampani ithi amamodeli asevele ekhiqiza iziphakamiso zokulungisa iningi lobuthakathaka obucutshungulwayo, okusekelwa iqiniso lokuthi i-Chrome ifaka imodeli ye-AI engu-4GB ukuze ithuthukise imisebenzi yendawo.
Ukuzenzakalelayo kwenqubo yokulungisa
I-AI ayigcini nje ngokuthola ubuthakathaka, kodwa futhi ihlanganyela kulo lonke umjikelezo wokuphila kobuthakathaka. Kusukela ekutholakaleni kuya ekwakhiweni kwama-patch, kufaka phakathi ukuhlukaniswa kwemibiko kanye nokwabiwa kwemisebenzi, uhlelo luhlunga ogaxekile kanye nama-duplicate, luphinda izivivinyo zobufakazi bomqondo, lunikeze amazinga obunzima, futhi luhambisa umbiko ngamunye eqenjini elinomthwalo wemfanelo . I-Google ilinganisela ukuthi le nqubo isindisa amakhulu amahora omsebenzi wonjiniyela bayo inyanga ngayinye.
Uma i-patch isilungile, ihlanganiswa ngqo kusuka endaweni yokugcina eyinhloko iye egatsheni elizinzile le-Chrome, eliqashwa njalo ukuze kuvinjelwe ubuthakathaka obusha. Le nkampani isiqalile nokwenza ngokuzenzakalelayo ukukhiqizwa kwamanothi okukhululwa kanye nezincazelo ze-CVE , isusa izithiyo ezenziwe ngesandla. NgoMeyi, lolu hlelo luvimbele ubuthakathaka obungaphezu kuka-20 ekufinyeleleni ekukhiqizeni, okuhlanganisa nobodwa obubhekwa njengobubalulekile.
Izibuyekezo ezisheshayo kanye nokulungiswa okunamandla
Ukuthola nokulungisa ubungozi akubavikeli abasebenzisi ngokushesha. Uma ukuguqulwa kuvela kukhodi yomthombo ovulekile, umhlaseli angakufunda ngaphambi kokuba i-patch ifinyelele kuwo wonke amakhompyutha. Ukuze kuncishiswe leli windi lokudalulwa, i-Google Chrome ishintsha ijubane layo futhi ibuyekeza njalo emavikini amabili ngezibuyekezo zokuphepha zamasonto onke, futhi isivele ihlola ukukhipha izibuyekezo zokuphepha ezimbili ngesonto. Ngaphezu kwalokho, isebenza ohlelweni lokulungisa olunamandla oluzovumela izibuyekezo ukuthi zisetshenziswe ngaphandle kwesidingo sokuqala kabusha isiphequluli.
Kusukela nge-Chrome 150, isiphequluli manje singaqala kabusha ngokuzenzakalelayo ku-macOS ukuze sifake isibuyekezo esilindile lapho sisebenza ngemuva ngaphandle kwamafasitela avuliwe. Umgomo wesikhathi eside ukugcina i-Chrome ivuselelwa njalo, ihlanganisa ukuphinda okuguquguqukayo, ukuqala kabusha okuzenzakalelayo ngezikhathi zokungasebenzi, kanye nokuvuselelwa kweseshini okuthuthukisiwe okuvimbela ukulahlekelwa amathebhu avulekile. I-Google iphinde ishintshele ezilimini eziphephile kwimemori njenge-Rust ukuze isuse zonke izinhlobo zobuthakathaka bokuphepha.
Le nkampani iqinisekisile ukuthi ithimba lezokuphepha le-Chrome laqala ukusebenzisa i-LLM ngo-2023, labambisana ne-DeepMind kanye ne-Project Zero kumathuluzi afana ne-Big Sleep, kwathi ekuqaleni kuka-2026 ladala i-ejenti esekelwe ku-Gemini eye yashintsha ukutholakala kobuthakathaka. Lo mkhuba awuhlukile ku-Google: I-Microsoft iphinde yalungisa ubuthakathaka obuyirekhodi obungu-570 ngoJuni ngenxa ye-AI , kuyilapho i-Apple igcina izibalo ezifanayo nezeminyaka edlule, okuphakamisa ukuthi ukuhlanganisa i-AI ekuphepheni kwenza umehluko omkhulu.
Izinga lokutholwa liye laphakama kakhulu kangangokuthi ithimba seliqalile ukukhipha ama-patch okuphepha kabili ngesonto, okuvame kakhulu kunokuvamile embonini. I-Google ilinganisela ukuthi ukuhlukaniswa okuzenzakalelayo kusindisa amakhulu amahora omsebenzi ngenyanga , yize ivuma ukuthi lokhu kulinganisa akuqondile. Le nkampani igcizelela ukuthi abantu basahilelekile: ama-patch akhiqizwe yi-AI athunyelwa kubathuthukisi ukuze bawabuyekeze futhi bawavume okokugcina, futhi uhlelo aluhlanganisi ama-patch ngokuzenzakalelayo.
Ukutholakala kwephutha elineminyaka engu-13 ubudala kubonisa amandla e-AI okuthola izinkinga ezingatholakali ngezindlela zendabuko. I-Google ihlela ukuqhubeka nokusebenzisa i-AI ukuvimbela amaphutha amasha ngokuwabona nokuwasusa eduze ngangokunokwenzeka ngesikhathi sokuthunyelwa kwekhodi. Le nkampani iphinde inciphise izinkinga esihlahleni sobuthakathaka ezingathinta ikhodi ebhekwa njengephephile uma ihlukanisiwe, isu lokuvimbela ubungozi obuhambisanayo.
Kubasebenzisi, isincomo esisebenzayo sihlala sifana: I-Chrome ilanda izibuyekezo zayo ngokuzenzakalelayo, kodwa kusadingeka ukuthi iqale kabusha ukuze isebenzise iningi lazo. Kunconywa ukuthi ulandele isaziso sokuvuselelwa bese uvala bese uvula kabusha isiphequluli uma kufaneleka. Ukuphepha kwe-Chrome kuthathe igxathu elikhulu phambili ngenxa ye-AI , kodwa ukuqapha okuqhubekayo kusadingeka endaweni lapho abahlaseli bangasebenzisa khona ubuchwepheshe obufanayo ukuthola ubuthakathaka.
