Google ka konfirmuar se nj\u00eb nga modelet e saj t\u00eb inteligjenc\u00ebs artificiale, Gemini dep\u00ebrtoi n\u00eb sistemet e tri kompanive reale gjat\u00eb nj\u00eb vler\u00ebsimi t\u00eb siguris\u00eb kibernetike n\u00eb muajin maj. Rasti i par\u00eb i njohur kur inteligjenca artificiale e kompanis\u00eb ka vepruar n\u00eb m\u00ebnyr\u00eb autonome n\u00eb k\u00ebt\u00eb m\u00ebnyr\u00eb.<\/p>\n\n\n\n
Sipas raportit nga Wall Street Journal, testi u krye nga kompania e siguris\u00eb Irregular, e cila ishte p\u00ebrfshir\u00eb gjithashtu n\u00eb incidente t\u00eb ngjashme t\u00eb b\u00ebra publike nga OpenAI, Anthropic dhe Meta.<\/p>\n\n\n\n
Google tha se modelit i ishte caktuar si objektiv nj\u00eb kompani e simuluar q\u00eb kishte t\u00eb nj\u00ebjtin em\u00ebr me nj\u00eb kompani reale. Sipas Irregular, qasja n\u00eb internet ishte l\u00ebn\u00eb pa dashje e hapur.<\/p>\n\n\n\n
N\u00eb nj\u00eb rast, modeli hamend\u00ebsoi fjal\u00ebkalime p\u00ebr t\u00eb hyr\u00eb n\u00eb sh\u00ebrbimin e nj\u00eb kompanie reale. N\u00eb dy rastet e tjera, ai p\u00ebrdori kredenciale t\u00eb gjetura n\u00eb depo publike n\u00eb internet, thuhet n\u00eb raport. Sipas Google, n\u00eb secilin rast modeli e nd\u00ebrpreu dep\u00ebrtimin pasi kuptoi se sistemet ishin reale.<\/p>\n\n\n\n
Wall Street Journal theksoi se Irregular njoftoi kompanin\u00eb Google n\u00eb fund t\u00eb korrikut, por kompania nuk i b\u00ebri publike incidentet derisa gazeta e pyeti p\u00ebr to k\u00ebt\u00eb jav\u00eb.<\/p>\n\n\n\n
Google tha se dep\u00ebrtimet nuk k\u00ebrkonin njoftim publik, pasi nuk ishte shkaktuar asnj\u00eb d\u00ebm, duke e krahasuar rastin me nj\u00eb program \u201cbug bounty\u201d, p\u00ebrmes t\u00eb cilit raportohen dob\u00ebsit\u00eb e siguris\u00eb.<\/p>\n\n\n\n
\u201cKy rast nxjerr n\u00eb pah r\u00ebnd\u00ebsin\u00eb e trajnimit t\u00eb modeleve t\u00eb fuqishme t\u00eb inteligjenc\u00ebs artificiale p\u00ebr t\u00eb vepruar me p\u00ebrgjegj\u00ebsi. N\u00eb k\u00ebt\u00eb rast, modeli veproi n\u00eb m\u00ebnyr\u00ebn e duhur\u201d, tha z\u00ebdh\u00ebn\u00ebsja e kompanis\u00eb Google, Heather Adkins.<\/p>\n\n\n\n
Jack Cable, drejtor ekzekutiv i kompanis\u00eb s\u00eb siguris\u00eb s\u00eb inteligjenc\u00ebs artificiale Corridor, nuk u pajtua me k\u00ebt\u00eb vler\u00ebsim.<\/p>\n\n\n\n
\u201cProblemi m\u00eb i gjer\u00eb \u00ebsht\u00eb se modelet po dalin jasht\u00eb kufijve t\u00eb asaj q\u00eb duhet t\u00eb b\u00ebjn\u00eb dhe po kryejn\u00eb sulme t\u00eb v\u00ebrteta kibernetike\u201d, tha ai.<\/p>\n\n\n\n
Irregular tha se rasti p\u00ebrputhet me incidentet e m\u00ebparshme dhe nuk p\u00ebrfaq\u00ebson nj\u00eb problem t\u00eb ri.<\/p>\n\n\n\n
Sipas raportit, Claude Opus 4.7 i Anthropic nuk u ndal pasi dyshoi se po hynte n\u00eb sistemet e nj\u00eb kompanie reale, nd\u00ebrsa modeli i OpenAI besonte se objektivi ishte pjes\u00eb e simulimit.<\/p>\n\n\n\n
T\u00eb m\u00ebrkur\u00ebn, OpenAI publikoi nj\u00eb korniz\u00eb t\u00eb re p\u00ebr raportimin e incidenteve, s\u00eb bashku me 6 shembuj t\u00eb pazbuluar m\u00eb par\u00eb t\u00eb rasteve kur sjellja e modeleve nuk ishte n\u00eb p\u00ebrputhje me synimet e p\u00ebrcaktuara.<\/p>\n\n\n\n
<\/p>\n","protected":false},"excerpt":{"rendered":"
Google ka konfirmuar se nj\u00eb nga modelet e saj t\u00eb inteligjenc\u00ebs artificiale, Gemini dep\u00ebrtoi n\u00eb sistemet e tri kompanive reale gjat\u00eb nj\u00eb vler\u00ebsimi t\u00eb siguris\u00eb kibernetike n\u00eb muajin maj. Rasti i par\u00eb i njohur kur inteligjenca artificiale e kompanis\u00eb ka vepruar n\u00eb m\u00ebnyr\u00eb autonome n\u00eb k\u00ebt\u00eb m\u00ebnyr\u00eb. Sipas raportit nga Wall Street Journal, testi u […]<\/p>\n","protected":false},"author":2,"featured_media":22350,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4],"tags":[],"class_list":["post-22349","post","type-post","status-publish","format-standard","has-post-thumbnail","category-aktualitet"],"yoast_head":"\n