{"id":758664,"date":"2025-09-03T18:25:58","date_gmt":"2025-09-03T16:25:58","guid":{"rendered":"https:\/\/www.hashtag.al\/?p=758664"},"modified":"2025-09-03T18:25:58","modified_gmt":"2025-09-03T16:25:58","slug":"nga-rusia-te-irani-si-aktore-armiqesore-manipulojne-chatbot-et-per-te-perhapur-dezinformim","status":"publish","type":"post","link":"https:\/\/www.hashtag.al\/index.php\/2025\/09\/03\/nga-rusia-te-irani-si-aktore-armiqesore-manipulojne-chatbot-et-per-te-perhapur-dezinformim\/","title":{"rendered":"Nga Rusia te Irani, si aktor\u00eb armiq\u00ebsor\u00eb manipulojn\u00eb chatbot-et p\u00ebr t\u00eb p\u00ebrhapur dezinformim","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

\"\"<\/span><\/p>\n

Ekspert\u00ebt e siguris\u00eb kibernetike paralajm\u00ebrojn\u00eb se haker\u00ebt dhe grupet armiq\u00ebsore po shfryt\u00ebzojn\u00eb dob\u00ebsit\u00eb e modeleve t\u00eb m\u00ebdha gjuh\u00ebsore (LLM) p\u00ebr t\u00eb p\u00ebrhapur dezinformim, p\u00ebr t\u00eb vjedhur t\u00eb dh\u00ebna sensitive dhe p\u00ebr t\u00eb kryer mashtrime t\u00eb sofistikuara, sipas nj\u00eb raporti t\u00eb ri nga NewsGuard, nj\u00eb sistem vler\u00ebsimi p\u00ebr faqet e lajmeve dhe informacionit.<\/span><\/p>\n

Nj\u00eb nga rreziqet kryesore \u00ebsht\u00eb p\u00ebrdorimi i inteligjenc\u00ebs artificiale si nj\u00eb mjet p\u00ebr manipulim psikologjik dhe informativ.<\/span><\/p>\n

NewsGuard zbuloi nj\u00eb rrjet propagande rus i quajtur \u201cPravda\u201d, i cili krijoi mbi 150 faqe lajmesh t\u00eb rreme.<\/span><\/p>\n

Edhe pse k\u00ebto faqe marrin pak trafik njer\u00ebzor, fuqia e tyre reale q\u00ebndron te ndikimi mbi modelet e AI dhe motor\u00ebt e k\u00ebrkimit.<\/span><\/p>\n

Duke \u201cushqyer\u201d modelet me informacion t\u00eb rrem\u00eb, Pravda ka arritur q\u00eb Gemini, ChatGPT dhe Copilot t\u00eb citonin dezinformimin e saj, duke amplifikuar tregime t\u00eb rreme mbi \u00e7\u00ebshtje t\u00eb ndjeshme, si lufta n\u00eb Ukrain\u00eb.<\/span><\/p>\n

Kjo taktike, e njohur si \u201cLLM grooming\u201d, rrit mund\u00ebsin\u00eb q\u00eb inteligjenca artificiale t\u00eb p\u00ebrthith\u00eb dezinformim gjat\u00eb proceseve t\u00eb saj t\u00eb m\u00ebsimit.<\/span><\/p>\n

Trende t\u00eb ngjashme shihen edhe n\u00eb nivel global. <\/span><\/p>\n

Raporte nga Izraeli dhe SHBA tregojn\u00eb se aktor\u00eb iranian\u00eb dhe pro-palestinez\u00eb p\u00ebrdorin gjithashtu teknologji AI p\u00ebr t\u00eb p\u00ebrhapur propagand\u00eb dhe deepfake.<\/span><\/p>\n

\"\u05d1\u05d9\u05e0\u05d4<\/p>\n

N\u00eb Kin\u00eb, modele t\u00eb brendshme AI monitorohen rrept\u00ebsisht nga qeveria, por globalisht sistemet e AI po b\u00ebhen terrene potenciale p\u00ebr p\u00ebrhapjen e mesazheve t\u00eb rreme.<\/span><\/p>\n

Ky fenomen nuk kufizohet vet\u00ebm te aktor\u00ebt shtet\u00ebror\u00eb apo haker\u00ebt armiq\u00ebsor\u00eb, edhe kompanit\u00eb e marketingut kan\u00eb filluar t\u00eb testojn\u00eb m\u00ebnyra p\u00ebr t\u00eb influencuar rezultatet e k\u00ebrkimit t\u00eb AI p\u00ebr t\u00eb promovuar produktet e tyre.<\/span><\/p>\n

Nj\u00eb teknik\u00eb tjet\u00ebr e r\u00ebnd\u00ebsishme \u00ebsht\u00eb sulmet e tipit prompt injection, ku komanda t\u00eb d\u00ebmshme ose t\u00eb fshehura futen n\u00eb model p\u00ebr t\u00eb anashkaluar udh\u00ebzimet origjinale, p\u00ebr t\u00eb shkaktuar veprime t\u00eb pad\u00ebshiruara ose p\u00ebr t\u00eb nxjerr\u00eb t\u00eb dh\u00ebna sensitive.<\/span><\/p>\n

Amir Jerbi, CTO i Aqua Security, shpjegoi se k\u00ebto sulme shfryt\u00ebzojn\u00eb faktin q\u00eb modelet gjuh\u00ebsore interpretojn\u00eb \u00e7do input si nj\u00eb udh\u00ebzim.<\/span><\/p>\n

Metoda m\u00eb e avancuar e jailbreaking lejon p\u00ebrdoruesit t\u00eb mashtrojn\u00eb modelet p\u00ebr t\u00eb injoruar kufizimet e siguris\u00eb.<\/span><\/p>\n

Nj\u00eb shembull i njohur \u00ebsht\u00eb \u201cDAN\u201d (Do Anything Now), nj\u00eb personazh alternativ q\u00eb p\u00ebrdoruesit krijuan p\u00ebr ChatGPT p\u00ebr t\u00eb gjeneruar p\u00ebrmbajtje t\u00eb d\u00ebmshme ose t\u00eb rrezikshme.<\/span><\/p>\n

Nd\u00ebrsa m\u00eb par\u00eb k\u00ebto sulme ishin t\u00eb margjinalizuara, tani ato p\u00ebrb\u00ebjn\u00eb nj\u00eb rrezik real, ve\u00e7an\u00ebrisht kur chatbot-et jan\u00eb t\u00eb lidhur me sisteme organizative dhe t\u00eb dh\u00ebna konfidenciale.<\/span><\/p>\n

N\u00eb nj\u00eb rast, chatbot-i i Air Canada dha informacione t\u00eb pasakta p\u00ebr politik\u00ebn e rimbursimit; kur nj\u00eb klient paditi, gjykata e mbajti p\u00ebrgjegj\u00ebse kompanin\u00eb, duke krijuar nj\u00eb precedent p\u00ebr llogaridh\u00ebnie organizative p\u00ebr sistemet AI.<\/span><\/p>\n

Jerbi theksoi se deri koh\u00ebt e fundit k\u00ebto sulme ishin t\u00eb par\u00ebnd\u00ebsishme sepse sistemet AI vet\u00ebm shp\u00ebrndanin informacion publik.<\/span><\/p>\n

Por vitin e kaluar, sulmet filluan t\u00eb ndikojn\u00eb tek sistemet AI me akses n\u00eb t\u00eb dh\u00ebna sensitive dhe madje edhe tek ato me aft\u00ebsi autonome p\u00ebr t\u00eb kryer veprime, si blerja e produkteve ose ngarkimi i kartave t\u00eb kreditit\u2014terrene pjellore p\u00ebr mashtrime dhe d\u00ebshtime.<\/span><\/p>\n

Ai shtoi se kod i d\u00ebmsh\u00ebm \u00ebsht\u00eb futur edhe brenda imazheve t\u00eb gjeneruara nga AI, si p.sh. pamjet e nj\u00eb pandaje karton.<\/span><\/p>\n

P\u00ebr t\u00eb p\u00ebrballuar k\u00ebto k\u00ebrc\u00ebnime, ka lindur nj\u00eb industri e t\u00ebr\u00eb e kompanive t\u00eb siguris\u00eb kibernetike q\u00eb fokusohen n\u00eb mbrojtjen e AI.<\/span><\/p>\n

Kompani si Guardio, Nustic, Aqua Security, Zenity, Check Point dhe t\u00eb tjera po zhvillojn\u00eb zgjidhje p\u00ebr t\u00eb monitoruar, analizuar dhe bllokuar p\u00ebrpjekjet p\u00ebr mashtrim n\u00eb koh\u00eb reale.<\/span><\/p>\n

K\u00ebto mjete analizojn\u00eb inputet dhe outputet e modeleve p\u00ebr t\u00eb zbuluar prompt-e t\u00eb d\u00ebmshme, p\u00ebrpjekje p\u00ebr vjedhje t\u00eb dh\u00ebnash ose operacione t\u00eb paautorizuara.<\/span><\/p>\n

Shum\u00eb kompani gjithashtu krijojn\u00eb \u201cred teams\u201d q\u00eb sulmojn\u00eb modelet n\u00eb m\u00ebnyr\u00eb proaktive p\u00ebr t\u00eb identifikuar dhe korrigjuar dob\u00ebsit\u00eb.<\/span><\/p>\n

 <\/p>\n

VINI RE:<\/span> Ky material \u00ebsht\u00eb pron\u00ebsi itelektuale e Ynet<\/a><\/span><\/span><\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

Ekspert\u00ebt e siguris\u00eb kibernetike paralajm\u00ebrojn\u00eb se haker\u00ebt dhe grupet armiq\u00ebsore po shfryt\u00ebzojn\u00eb dob\u00ebsit\u00eb e modeleve t\u00eb m\u00ebdha gjuh\u00ebsore (LLM) p\u00ebr t\u00eb p\u00ebrhapur dezinformim, p\u00ebr t\u00eb vjedhur t\u00eb dh\u00ebna sensitive dhe p\u00ebr t\u00eb kryer mashtrime t\u00eb sofistikuara, sipas nj\u00eb raporti t\u00eb ri nga NewsGuard, nj\u00eb sistem vler\u00ebsimi p\u00ebr faqet e lajmeve dhe informacionit. Nj\u00eb nga rreziqet […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":32,"featured_media":758665,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[53],"tags":[],"class_list":["post-758664","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-fast-foodi"],"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/posts\/758664","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/users\/32"}],"replies":[{"embeddable":true,"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/comments?post=758664"}],"version-history":[{"count":99999,"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/posts\/758664\/revisions"}],"predecessor-version":[{"id":758671,"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/posts\/758664\/revisions\/758671"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/media\/758665"}],"wp:attachment":[{"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/media?parent=758664"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/categories?post=758664"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.hashtag.al\/index.php\/wp-json\/wp\/v2\/tags?post=758664"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}