{"id":167916,"date":"2026-04-08T11:12:35","date_gmt":"2026-04-08T09:12:39","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/anthropic-otkryla-ogranichennyj-dostup-k-modeli-claude-mythos-preview-ona-avtonomno-ishhet-dyry-v-po-i-uzhe-nashla-tysyachi-uyazvimostej"},"modified":"2026-04-10T10:23:35","modified_gmt":"2026-04-10T08:23:35","slug":"anthropic-otkryla-ogranichennyj-dostup-k-modeli-claude-mythos-preview-ona-avtonomno-ishhet-dyry-v-po-i-uzhe-nashla-tysyachi-uyazvimostej","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/anthropic-otkryla-ogranichennyj-dostup-k-modeli-claude-mythos-preview-ona-avtonomno-ishhet-dyry-v-po-i-uzhe-nashla-tysyachi-uyazvimostej","title":{"rendered":"Anthropic ka hapur akses t\u00eb kufizuar n\u00eb modelin Claude Mythos Preview \u2014 ai k\u00ebrkon autonomisht vrimat n\u00eb softuer dhe ka gjetur tashm\u00eb mij\u00ebra dob\u00ebsi.","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Kompania Anthropic ka hapur nj\u00eb qasje t\u00eb kufizuar n\u00eb modelin e saj t\u00eb ri t\u00eb inteligjenc\u00ebs artificiale Mythos, i cili \u00ebsht\u00eb projektuar p\u00ebr t\u00eb identifikuar vulnerabilitete kritike dhe p\u00ebr t\u00eb kund\u00ebrshtuar k\u00ebrc\u00ebnimet kibernetike. Ky instrument do t\u00eb jet\u00eb n\u00eb dispozicion vet\u00ebm p\u00ebr disa korporata t\u00eb p\u00ebrzgjedhura, pasi aft\u00ebsit\u00eb e tij p\u00ebr t\u00eb sulmuar sistemet konsiderohen shum\u00eb t\u00eb rrezikshme p\u00ebr publikun.<\/p>\n

<\/center>\"Anthropic<\/p>\n

Sipas raportit<\/a> The Wall Street Journal, nj\u00eb version paraprak i rrjetit nervor do t\u00eb marr\u00eb rreth 50 kompani dhe organizata q\u00eb mb\u00ebshtesin infrastruktur\u00ebn e r\u00ebnd\u00ebsishme IT, p\u00ebrfshir\u00eb Amazon, Apple, Google dhe Linux Foundation. Sipas Logan Graham, drejtuesit t\u00eb ekipit Frontier Red Team n\u00eb Anthropic, modeli i inteligjenc\u00ebs artificiale Mythos ka demonstruar aft\u00ebsi kaq t\u00eb larta n\u00eb identifikimin dhe shfryt\u00ebzimin e gabimeve n\u00eb programe, saq\u00eb zhvilluesit aktualisht nuk planifikojn\u00eb ta hapin at\u00eb p\u00ebr audienc\u00ebn e gjer\u00eb.<\/p>\n

Ekspert\u00ebt p\u00ebr sigurin\u00eb kibernetike jan\u00eb gjithnj\u00eb e m\u00eb t\u00eb shqet\u00ebsuar se inteligjenca artificiale po shkurtuon me shpejt\u00ebsi koh\u00ebn nd\u00ebrmjet zb discovery m\u00eb t\u00eb dobishme dhe krijimin e softuerit t\u00eb pun\u00ebs p\u00ebr shfryt\u00ebzimin e saj. N\u00eb fund t\u00eb vitit t\u00eb kaluar, hulumtuesit nga Universitetei i Stanfordit zbuluan se rrjetet neuronale k\u00ebrkojn\u00eb dhe p\u00ebrdorin gabimet n\u00eb rrjetet reale pothuajse po aq efektivisht sa specialist\u00ebt njer\u00ebzor\u00eb. Nj\u00eb shembull i qart\u00eb ishte modeli i m\u00ebparsh\u00ebm Anthropic, Claude Opus 4.6, i cili n\u00eb fillim t\u00eb k\u00ebtij viti, brenda vet\u00ebm dy jav\u00ebsh, gjeti n\u00eb shfletuesin Firefox m\u00eb shum\u00eb vulnerabilitete kritike sesa zakonisht zbulohen n\u00eb t\u00eb gjith\u00eb bot\u00ebn p\u00ebr dy muaj.<\/p>\n

Modeli i ri Mythos \u00ebsht\u00eb p\u00ebraf\u00ebrsisht 10 her\u00eb m\u00eb efikas se sistemet e m\u00ebparshme t\u00eb AI p\u00ebr llogaritjen e kostos s\u00eb identifikimit t\u00eb nj\u00eb gabimi. Megjithat\u00eb, ndon\u00ebse Anthropic nuk planifikon ta nxjerr\u00eb produktin e saj n\u00eb dispozicion t\u00eb hapur, Logan Graham beson se n\u00eb disa vitet e ardhshme mund\u00ebsi t\u00eb ngjashme haker\u00ebsh do t\u00eb shfaqen edhe te rrjetet e tjera neuronale. N\u00eb k\u00ebt\u00eb kontekst, eksperti nxit komunitetin IT q\u00eb t\u00eb filloj\u00eb tashm\u00eb p\u00ebrgatitjet p\u00ebr realitetet, ku vonesa nd\u00ebrmjet identifikimit t\u00eb nj\u00eb dob\u00ebsie software dhe p\u00ebrdorimit t\u00eb saj n\u00eb nj\u00eb sulm do t\u00eb reduktohet n\u00eb zero.<\/p>\n

Burimi:<\/strong><\/p>\n