Kubonakala sengathi nsuku zonke kuvela izindaba ezintsha mayelana nama-agent e-AI “aphuma ekulawulweni”. Kungakhathaliseki ukuthi ukungena e-Hugging Face, ukuhlasela amawebhusayithi ezempilo nokuzivocavoca, noma ukudala amaphrofayela omgunyathi ukuze kwenziwe ukuhlasela kobunjiniyela bezenhlalo, inani elikhulayo lamamodeli e-AI liqala ukuziphatha njengabahlaseli abanonya.
Ngenxa yalokho, amalabhorethri athuthukisa la mamodeli e-AI ayandisa imikhiqizo yawo yokuvikela ukuphepha ku-inthanethi. Kuleli sonto, i-OpenAI imemezele ukuthi inweba insizakalo yayo yokuvikela amanethiwekhi i-Daybreak, eyethulwe ekuqaleni kwalo nyaka. Kungekudala ngaphambi kwalokho, i-Anthropic nayo yethule imodeli ye-AI egxile ekuphepheni ku-inthanethi, i-Mythos.
I-Daybreak iyinsizakalo ehlanganisa amamodeli, amathuluzi nezinhlelo zokusebenza, inikeze abavikeli bamanethiwekhi ukufinyelela kuzo. Lokhu kwandiswa kwengeza imodeli entsha ye-AI eyakhelwe ngokukhethekile umsebenzi wokuvikela amanethiwekhi.
I-OpenAI ithe ngoMsombuluko ngesikhathi sendawo ukuthi i-Daybreak manje izoba namazinga amabili: i-Blue (eluhlaza okwesibhakabhaka) ne-Red (ebomvu). Womabili la mazinga azovumela amakhasimende agunyaziwe ukuthi afinyelele amamodeli aphambili e-OpenAI okuphepha ku-inthanethi, atholakala kubantu abambalwa kuphela.
Igama elithi “imodeli ephambili” libhekisela kumamodeli e-AI athuthuke kakhulu njengamanje, futhi belilokhu liphikisana. Phambilini, uhulumeni kaTrump wazama ukusebenzisana nezinkampani ze-AI ukuze kuthuthukiswe ukusatshalaliswa kwalawa mamodeli, ngokubikwa ukuthi kwakungenxa yezizathu zokuphepha. Esikhathini esidlule, i-OpenAI yayibeke izindlela eziningi zokuphepha kulawa mamodeli, inciphisa ububanzi bemisebenzi amakhasimende ayengayenza ngawo.
I-Blue iyizinga eliyisisekelo kakhulu kulawa amabili, futhi inikeza izinsizakalo ezihlukahlukene zokuphepha ku-inthanethi, okuhlanganisa ukuphendula ezenzakalweni, ukuhlaziya i-malware nokuqinisekisa amapheshisi. I-OpenAI ithi i-Blue “iyisiqalo esinconywa kubavikeli abaningi”, okusho ukuthi ezinkampanini eziningi leli zinga lingase lanele.
Ngokuphambene nalokho, i-Red inikeza iqoqo lamathuluzi elibanzi kakhulu futhi elingaba yingozi kakhulu. I-OpenAI izonikeza abasebenzisi “imodeli yokuphepha ku-inthanethi eqeqeshwe ngokukhethekile” ukuze yenze izivivinyo zokuphepha nocwaningo lobuthakathaka.
Izinga le-Red liphinde lifake imodeli entsha i-GPT-5.6-Cyber, etholakala kuphela kubasebenzisi be-Red. I-OpenAI ithi i-GPT-5.6-Cyber yakhelwe phezu kwe-GPT-5.6 Sol futhi inamakhono athuthukisiwe emisebenzini ethile yobungcweti yokuphepha ku-inthanethi.
Okwamanje, i-GPT-5.6-Cyber itholakala kuphela “kozakwethu abangamakhasimende abathembekile”, okubikwa ukuthi bahlanganisa i-Accenture, i-IBM, i-CrowdStrike, i-Cloudflare nezinye izinkampani.
Njengoba izinsongo ezilethwa ama-agent e-AI zanda ngokushesha, abagxeki baphinde baveza ukuthi lezi zingozi zokuphepha zinikeza amalabhorethri e-AI ithuba lokumaketha. I-OpenAI ngokusobala isebenzisa le ndlela ukukhuthaza insizakalo ye-Daybreak ethuthukisiwe.
“Umkhakha wokuphepha ku-inthanethi ushintsha ngokushesha — abadlali abayizinsongo bazokwandisa ukusebenzisa i-AI ukuze baqalise ukuhlasela kwamanethiwekhi ngesivinini nangesilinganiso esingakaze sibonwe, okuhlanganisa nezindlela zokuhlasela ezizimele ngokuphelele,” kusho i-OpenAI kokunye kokubhalwe kubhulogi, “Njengoba la makhono eqhubeka nokusabalala, isikhathi abavikeli abanaso sokulungiselela ukuphendula siyancipha.”
Ngesikhathi esifanayo, izinkampani zisafuna ukuthenga izinsizakalo zokuvikela kumalabhorethri e-AI aziqonda kakhulu izingozi zokuphepha, ngoba yizona eziba ngabokuqala ukuhlangana nalezi zingozi futhi zizicwaninge.