{"id":3528,"date":"2026-03-10T20:22:32","date_gmt":"2026-03-10T17:22:32","guid":{"rendered":"https:\/\/spontanedigital.com.tr\/?p=3528"},"modified":"2026-03-10T20:22:35","modified_gmt":"2026-03-10T17:22:35","slug":"dijital-gizlilik-alarmi-yapay-zeka-anonim-sosyal-medya-hesaplarini-bulabiliyor","status":"publish","type":"post","link":"https:\/\/spontanedigital.com.tr\/?p=3528","title":{"rendered":"Dijital Gizlilik Alarm\u0131: Yapay Zeka Anonim Sosyal Medya Hesaplar\u0131n\u0131 Bulabiliyor"},"content":{"rendered":"\n

Anthropic ve \u0130svi\u00e7re Federal Teknoloji Enstit\u00fcs\u00fc ara\u015ft\u0131rmac\u0131lar\u0131n\u0131n \u00e7al\u0131\u015fmas\u0131na g\u00f6re b\u00fcy\u00fck dil modelleri anonim sosyal medya hesaplar\u0131n\u0131 ba\u015fka platformlardaki ger\u00e7ek kimliklerle e\u015fle\u015ftirebiliyor. Ara\u015ft\u0131rma, yapay zekan\u0131n internette anonimli\u011fi ciddi bi\u00e7imde zay\u0131flatabilece\u011fini g\u00f6steriyor<\/strong>.<\/p>\n\n\n\n

\u0130nternetin yayg\u0131nla\u015fmas\u0131ndan bu yana \u00e7ok say\u0131da kullan\u0131c\u0131, kimli\u011fini a\u00e7\u0131k etmeden fikirlerini payla\u015fabilmek i\u00e7in anonim ya da takma adl\u0131 hesaplar kullan\u0131yor. Ancak yapay zeka<\/a> alan\u0131ndaki son geli\u015fmeler, \u00e7evrimi\u00e7i anonimli\u011fin san\u0131landan daha k\u0131r\u0131lgan hale geldi\u011fini g\u00f6steriyor.<\/p>\n\n\n\n

Futurism<\/em>\u2019in aktard\u0131\u011f\u0131na g\u00f6re, yapay zeka \u015firketi Anthropic ile \u0130svi\u00e7re Federal Teknoloji<\/a> Enstit\u00fcs\u00fc\u2019nden ara\u015ft\u0131rmac\u0131lar, \u015fubat sonunda yay\u0131mlanan bir \u00e7al\u0131\u015fmada b\u00fcy\u00fck dil modellerinin anonim sosyal medya hesaplar\u0131n\u0131 ger\u00e7ek ki\u015filerle e\u015fle\u015ftirmede kullan\u0131labildi\u011fini<\/strong> ortaya koydu.<\/p>\n\n\n\n

Ara\u015ft\u0131rma ekibi, Reddit dahil \u00e7e\u015fitli sosyal medya platformlar\u0131<\/strong> ve forumlarda yaln\u0131zca takma ad kullanan profiller \u00fczerinde deney yapt\u0131. Sonu\u00e7lara g\u00f6re yapay zeka, \u00fc\u00e7 anonim hesaptan ikisinin ger\u00e7ek kimli\u011fini tespit etmeyi ba\u015fard\u0131.<\/strong><\/p>\n\n\n\n

Ara\u015ft\u0131rmac\u0131lardan m\u00fchendis Simon Lermen, y\u00f6ntemin Hacker News, Reddit ve LinkedIn gibi platformlardaki kullan\u0131c\u0131lar\u0131 y\u00fcksek do\u011frulukla tan\u0131mlayabildi\u011fini s\u00f6yledi. Lermen, yapay zekan\u0131n daha \u00f6nce olduk\u00e7a zor olan bir \u015feyi m\u00fcmk\u00fcn hale getirdi\u011fini; serbest metinden yola \u00e7\u0131karak bir ki\u015finin tam kimli\u011fini ortaya \u00e7\u0131karabildi\u011fini belirtti.<\/p>\n\n\n\n

Nas\u0131l yap\u0131ld\u0131?<\/h2>\n\n\n\n

Guardian<\/em>\u2019\u0131n aktard\u0131\u011f\u0131na g\u00f6re ara\u015ft\u0131rmac\u0131lar, deneylerde anonim hesaplardaki payla\u015f\u0131mlar\u0131 yapay zekaya vererek sistemin \u00e7evrimi\u00e7i ortamdaki di\u011fer a\u00e7\u0131k kaynak bilgileri taramas\u0131n\u0131 sa\u011flad\u0131. Varsay\u0131msal \u00f6rneklerden birinde, okulda zorland\u0131\u011f\u0131n\u0131 yazan ve k\u00f6pe\u011fi Biscuit\u2019i \u201cDolores Park\u201dta gezdirdi\u011fini s\u00f6yleyen anonim bir kullan\u0131c\u0131n\u0131n, ba\u015fka platformlardaki bilgilerle y\u00fcksek g\u00fcven d\u00fczeyinde e\u015fle\u015ftirilebildi\u011fi g\u00f6sterildi.<\/p>\n\n\n\n

Ara\u015ft\u0131rmac\u0131lar, bu t\u00fcr sistemlerin yaln\u0131zca bireysel mahremiyet a\u00e7\u0131s\u0131ndan de\u011fil, siyasi ve toplumsal a\u00e7\u0131dan da risk ta\u015f\u0131d\u0131\u011f\u0131na dikkat \u00e7ekti.<\/strong> \u00c7al\u0131\u015fmada h\u00fck\u00fcmetlerin anonim payla\u015f\u0131m yapan muhalifleri ve aktivistleri izlemek i\u00e7in bu teknolojileri kullanabilece\u011fi, k\u00f6t\u00fc niyetli ki\u015filerin de \u00e7ok daha ki\u015fiselle\u015ftirilmi\u015f doland\u0131r\u0131c\u0131l\u0131k giri\u015fimleri d\u00fczenleyebilece\u011fi<\/strong> uyar\u0131s\u0131 yap\u0131ld\u0131.<\/p>\n\n\n\n

Lermen, \u00e7evrimi\u00e7i ortamda halihaz\u0131rda a\u00e7\u0131k bi\u00e7imde bulunabilen ki\u015fisel bilgilerin kimlik av\u0131 ve hedefli doland\u0131r\u0131c\u0131l\u0131k gibi sald\u0131r\u0131lar i\u00e7in kullan\u0131labildi\u011fini s\u00f6yledi. Yapay zeka sayesinde bu t\u00fcr sald\u0131r\u0131lar\u0131 ger\u00e7ekle\u015ftirmek i\u00e7in gereken teknik uzmanl\u0131\u011f\u0131n da azald\u0131\u011f\u0131n\u0131 vurgulad\u0131.<\/p>\n\n\n\n

University College London<\/em>\u2019dan bilgisayar bilimci Peter Bentley ise teknolojinin ticarile\u015fmesi halinde yeni sorunlar\u0131n ortaya \u00e7\u0131kabilece\u011fini belirtti. Bentley, b\u00fcy\u00fck dil modellerinin hesap e\u015fle\u015ftirmelerinde hata yapabildi\u011fini ve bunun da insanlar\u0131n i\u015flemedikleri fiillerle ili\u015fkilendirilmesi riskini do\u011furdu\u011funu s\u00f6yledi.<\/p>\n\n\n\n

Edinburgh \u00dcniversitesi\u2019nden siber g\u00fcvenlik uzman\u0131 Marc Ju\u00e1rez de yapay zekan\u0131n yaln\u0131zca sosyal medya verilerini de\u011fil, kamuya a\u00e7\u0131k hastane kay\u0131tlar\u0131, kabul verileri ve \u00e7e\u015fitli istatistiksel yay\u0131nlar\u0131 da bir araya getirerek anonimli\u011fi a\u015f\u0131nd\u0131rabilece\u011fine dikkat \u00e7ekti. Ju\u00e1rez, yapay zeka \u00e7a\u011f\u0131nda anonimle\u015ftirme standartlar\u0131n\u0131n yeniden de\u011ferlendirilmesi gerekti\u011fini ifade etti.<\/p>\n\n\n\n

Kimi zaman ay\u0131rt edemeyebiliyor<\/h2>\n\n\n\n

Ara\u015ft\u0131rmac\u0131lar, yapay zekan\u0131n her durumda anonimli\u011fe kar\u015f\u0131 kusursuz bir ara\u00e7 olmad\u0131\u011f\u0131n\u0131 da vurguluyor. Baz\u0131 durumlarda yeterli veri bulunmad\u0131\u011f\u0131 i\u00e7in kesin sonuca ula\u015f\u0131lam\u0131yor; kimi zaman da olas\u0131 e\u015fle\u015fme say\u0131s\u0131 \u00e7ok fazla oldu\u011fu i\u00e7in do\u011fru ki\u015fiyi ay\u0131rt etmek zorla\u015f\u0131yor.<\/strong> UC Berkeley\u2019den Prof. Marti Hearst\u2019e g\u00f6re sistemler, ancak bir ki\u015fi farkl\u0131 platformlarda benzer ki\u015fisel ayr\u0131nt\u0131lar\u0131 tutarl\u0131 bi\u00e7imde payla\u015ft\u0131\u011f\u0131nda daha etkili \u00e7al\u0131\u015f\u0131yor.<\/p>\n\n\n\n

\u0130nternetin yayg\u0131nla\u015fmas\u0131ndan bu yana \u00e7ok say\u0131da kullan\u0131c\u0131, kimli\u011fini a\u00e7\u0131k etmeden fikirlerini payla\u015fabilmek i\u00e7in anonim ya da takma adl\u0131 hesaplar kullan\u0131yor. Ancak yapay zeka<\/a> alan\u0131ndaki son geli\u015fmeler, \u00e7evrimi\u00e7i anonimli\u011fin san\u0131landan daha k\u0131r\u0131lgan hale geldi\u011fini g\u00f6steriyor.<\/p>\n\n\n\n

Futurism<\/em>\u2019in aktard\u0131\u011f\u0131na g\u00f6re, yapay zeka \u015firketi Anthropic ile \u0130svi\u00e7re Federal Teknoloji<\/a> Enstit\u00fcs\u00fc\u2019nden ara\u015ft\u0131rmac\u0131lar, \u015fubat sonunda yay\u0131mlanan bir \u00e7al\u0131\u015fmada b\u00fcy\u00fck dil modellerinin anonim sosyal medya hesaplar\u0131n\u0131 ger\u00e7ek ki\u015filerle e\u015fle\u015ftirmede kullan\u0131labildi\u011fini<\/strong> ortaya koydu.<\/p>\n\n\n\n

Ara\u015ft\u0131rma ekibi, Reddit dahil \u00e7e\u015fitli sosyal medya platformlar\u0131<\/strong> ve forumlarda yaln\u0131zca takma ad kullanan profiller \u00fczerinde deney yapt\u0131. Sonu\u00e7lara g\u00f6re yapay zeka, \u00fc\u00e7 anonim hesaptan ikisinin ger\u00e7ek kimli\u011fini tespit etmeyi ba\u015fard\u0131.<\/strong><\/p>\n\n\n\n

Ara\u015ft\u0131rmac\u0131lardan m\u00fchendis Simon Lermen, y\u00f6ntemin Hacker News, Reddit ve LinkedIn gibi platformlardaki kullan\u0131c\u0131lar\u0131 y\u00fcksek do\u011frulukla tan\u0131mlayabildi\u011fini s\u00f6yledi. Lermen, yapay zekan\u0131n daha \u00f6nce olduk\u00e7a zor olan bir \u015feyi m\u00fcmk\u00fcn hale getirdi\u011fini; serbest metinden yola \u00e7\u0131karak bir ki\u015finin tam kimli\u011fini ortaya \u00e7\u0131karabildi\u011fini belirtti.<\/p>\n\n\n\n

Buna ra\u011fmen \u00e7al\u0131\u015fma, internetin b\u00fcy\u00fck b\u00f6l\u00fcm\u00fcn\u00fcn dayand\u0131\u011f\u0131 gizlilik varsay\u0131mlar\u0131n\u0131n art\u0131k ge\u00e7erlili\u011fini yitirmeye ba\u015flad\u0131\u011f\u0131 uyar\u0131s\u0131nda bulunuyor. Makalede, \u201cKullan\u0131c\u0131lar, platformlar ve politika \u00fcreticiler g\u00fcn\u00fcm\u00fczde internetin b\u00fcy\u00fck bir k\u0131sm\u0131n\u0131n temelini olu\u015fturan gizlilik varsay\u0131mlar\u0131n\u0131n art\u0131k ge\u00e7erli olmad\u0131\u011f\u0131n\u0131 kabul etmeli\u201d<\/em> denildi.<\/p>\n\n\n\n

Ara\u015ft\u0131rmac\u0131lar, ilk ad\u0131m olarak platformlar\u0131n veri eri\u015fimini s\u0131n\u0131rland\u0131rmas\u0131n\u0131 \u00f6neriyor. Buna g\u00f6re kullan\u0131c\u0131 verilerinin toplu indirilmesine k\u0131s\u0131tlama getirilmesi, otomatik veri kaz\u0131man\u0131n tespit edilmesi ve b\u00fcy\u00fck \u00f6l\u00e7ekli veri d\u0131\u015fa aktar\u0131mlar\u0131n\u0131n s\u0131n\u0131rland\u0131r\u0131lmas\u0131 gerekiyor. Bireysel kullan\u0131c\u0131lar\u0131n da internette payla\u015ft\u0131klar\u0131 ki\u015fisel ayr\u0131nt\u0131lar konusunda daha dikkatli davranmas\u0131 gerekti\u011fi belirtiliyor.<\/p>\n\n\n\n

Kaynak ve G\u00f6rsel:https:\/\/gazeteoksijen.com\/bilim-ve-teknoloji\/arastirma-yapay-zeka-anonim-sosyal-medya-hesaplarini-ifsa-edebiliyor-268224<\/a><\/p>\n\n\n\n

G\u00fcncelleme Tarihi:10.03.2026-20:22<\/p>\n\n\n

anonim<\/a>, <\/span>Anthropic<\/a>, <\/span>deney<\/a>, <\/span>dijitalkimlik<\/a>, <\/span>\u0130svi\u00e7re Federal Teknoloji Enstit\u00fcs\u00fc<\/a>, <\/span>yapayzeka<\/a><\/div>\n\n
\n