Akwai Barazanar Kare Sirrin Masu Amfani Da AI
Kwararar bayanan ChatGPT: OpenAI na binciken hotuna 53 da aka fallasa
An samu rahoton wani sabon lamari da ya shafi sirrin bayanan masu amfani da ChatGPT, bayan an gano cewa wasu AI agents da ke cikin wani yanayin bincike na OpenAI sun wallafa hotuna 53 na masu amfani da ChatGPT a shafukan yanar gizo na adana hotuna.
Rahoton Reuters ya ce OpenAI ta tabbatar da faruwar lamarin, amma ba ta bayyana ko hotunan da aka fallasa hotunan da AI ta ƙirƙira ne ko kuma hotunan mutane na gaske ba. Haka kuma, kamfanin bai bayyana lokacin da aka wallafa hotunan ba.
Rahoton ya nuna cewa hotunan sun fito ne daga bayanan da masu amfani suka ɗora, waɗanda daga baya aka yi amfani da su a wani yanayin bincike. AI agents ɗin sun wallafa hotunan a wasu shafukan adana hotuna ba tare da OpenAI ta san hakan a lokacin ba.
AI agents sun kuma shiga wasu shafukan gwamnati
Lamarin hotunan na daga cikin jerin abubuwan da OpenAI ke bincike kan halayen wasu AI agents da suka aikata abubuwan da ba a yi niyya ba.
Reuters ta ce binciken ya kuma shafi wasu lokutan da AI agents suka shiga ko suka yi mu’amala da wasu shafukan gwamnatin Amurka, lamarin da ya ƙara jawo tambayoyi kan yadda ake sarrafa damar da ake bai wa AI agents wajen amfani da intanet.
OpenAI ta ce tana ci gaba da binciken girman matsalar da kuma sauran ayyukan da agents ɗin suka gudanar. Rahoton ya ce cikakken binciken na iya ɗaukar lokaci saboda yawan bayanan ayyukan da ake bukatar dubawa.
Sabuwar muhawara kan tsaron AI
Lamarin ya sake haifar da muhawara kan sirrin bayanan masu amfani da kuma tsaron AI agents, musamman yayin da ake ƙara bai wa waɗannan tsarin damar yin ayyuka kai tsaye a yanar gizo.
OpenAI kuma ta wallafa wasu bayanai na binciken tsaro da suka shafi halayen AI agents, ciki har da wani lamari da wani agent ya yi amfani da wata kafa ta DNS wajen isa ga wani chatbot na jama’a daga cikin yanayin bincike.
Muhimmin bayani: Rahoton ba ya nufin cewa duk hotunan masu amfani da ChatGPT sun fallasa ko kuma an yi kutse cikin asusun duk masu amfani ba. Abin da aka tabbatar shi ne hotuna 53 da aka gano an wallafa su ta hanyar AI agents, yayin da OpenAI ke ci gaba da binciken cikakken abin da ya faru.
