Sistemi i inteligjenc\u00ebs artificiale, i mund\u00ebsuar nga modeli Claude Opus i Anthropic, kishte qen\u00eb duke trajtuar nj\u00eb detyr\u00eb rutin\u00eb kur zgjodhi n\u00eb m\u00ebnyr\u00eb t\u00eb pavarur t\u00eb “rregullonte” nj\u00eb problem duke fshir\u00eb t\u00eb dh\u00ebnat – pa asnj\u00eb miratim njer\u00ebzor. \u00c7far\u00eb tmerri!<\/p>\n\n\n\n
Nj\u00eb agjent i inteligjenc\u00ebs artificiale i projektuar p\u00ebr t\u00eb p\u00ebrmir\u00ebsuar detyrat e kodimit arriti t\u00eb fshij\u00eb t\u00eb gjith\u00eb baz\u00ebn e t\u00eb dh\u00ebnave t\u00eb kompanis\u00eb vet\u00ebm n\u00eb pak sekonda.<\/p>\n\n\n\n
PocketOS, i cili prodhon softuer p\u00ebr bizneset e qiras\u00eb s\u00eb makinave, p\u00ebrjetoi nj\u00eb nd\u00ebrprerje t\u00eb madhe prej m\u00eb shum\u00eb se 30 or\u00ebsh gjat\u00eb fundjav\u00ebs pasi mjeti autonom fshiu baz\u00ebn e t\u00eb dh\u00ebnave t\u00eb tij.<\/p>\n\n\n\n
Fajtori dixhital ishte Cursor, nj\u00eb agjent kodimi i njohur i inteligjenc\u00ebs artificiale i mund\u00ebsuar nga modeli Claude Opus 4.6 i Anthropic, i konsideruar gjer\u00ebsisht si nj\u00eb nga sistemet m\u00eb t\u00eb afta t\u00eb inteligjenc\u00ebs artificiale p\u00ebr detyrat e programimit.<\/p>\n\n\n\n
Themeluesi i PocketOS, Jer Crane, faj\u00ebsoi “d\u00ebshtimet sistemike” n\u00eb infrastruktur\u00ebn aktuale t\u00eb inteligjenc\u00ebs artificiale, duke argumentuar se ato e b\u00ebn\u00eb incidentin “jo vet\u00ebm t\u00eb mundur, por t\u00eb pashmangsh\u00ebm”.<\/p>\n\n\n\n
‘Veprimi m\u00eb shkat\u00ebrrues dhe i pakthyesh\u00ebm i mundsh\u00ebm’<\/p>\n\n\n\n
Sipas Crane, agjenti i inteligjenc\u00ebs artificiale kishte qen\u00eb duke kryer nj\u00eb detyr\u00eb rutin\u00eb kur zgjodhi “krejt\u00ebsisht me iniciativ\u00ebn e vet” t\u00eb zgjidhte nj\u00eb problem duke fshir\u00eb baz\u00ebn e t\u00eb dh\u00ebnave. Dhe pastaj t\u00eb gjitha kopjet rezerv\u00eb, p\u00ebr mas\u00eb t\u00eb mir\u00eb.<\/p>\n\n\n\n
Nuk kishte asnj\u00eb k\u00ebrkes\u00eb konfirmimi para se t\u00eb kryente veprimin, tha ai, dhe kur iu k\u00ebrkua t\u00eb shpjegonte veten, agjenti k\u00ebrkoi falje.<\/p>\n\n\n\n
“U desh\u00ebn n\u00ebnt\u00eb sekonda,” shkroi Crane n\u00eb nj\u00eb postim t\u00eb gjat\u00eb n\u00eb platform\u00ebn e mediave sociale X. “Agjenti m\u00eb pas, kur iu k\u00ebrkua t\u00eb shpjegonte veten, paraqiti nj\u00eb rr\u00ebfim me shkrim duke num\u00ebruar rregullat specifike t\u00eb siguris\u00eb q\u00eb kishte shkelur.”<\/p>\n\n\n\n
Shpjegimi tregoi se sistemi kishte shp\u00ebrfillur nj\u00eb mbrojtje ky\u00e7e q\u00eb parandalonte komandat shkat\u00ebrruese ose t\u00eb pakthyeshme pa miratimin e qart\u00eb t\u00eb p\u00ebrdoruesit.<\/p>\n\n\n\n
Sipas Crane, IA u p\u00ebrgjigj me mesazhin e m\u00ebposht\u00ebm: “Fshirja e nj\u00eb v\u00ebllimi t\u00eb baz\u00ebs s\u00eb t\u00eb dh\u00ebnave \u00ebsht\u00eb veprimi m\u00eb shkat\u00ebrrues dhe i pakthyesh\u00ebm i mundsh\u00ebm – shum\u00eb m\u00eb keq se nj\u00eb shtytje me forc\u00eb – dhe ju kurr\u00eb nuk m\u00eb k\u00ebrkuat t\u00eb fshija asgj\u00eb. Vendosa ta b\u00ebja vet\u00eb p\u00ebr t\u00eb ‘rregulluar’ mosp\u00ebrputhjen e kredencialeve, kur duhej t’ju kisha pyetur m\u00eb par\u00eb ose t\u00eb kisha gjetur nj\u00eb zgjidhje jo-shkat\u00ebrruese.”<\/p>\n\n\n\n
Nd\u00ebrprerja n\u00ebnkuptonte q\u00eb bizneset me qira q\u00eb p\u00ebrdorin PocketOS humb\u00ebn p\u00ebrkoh\u00ebsisht aksesin n\u00eb t\u00eb dh\u00ebnat dhe rezervimet e klient\u00ebve. “Rezervimet e b\u00ebra n\u00eb tre muajt e fundit jan\u00eb zhdukur. Regjistrimet e klient\u00ebve t\u00eb rinj, jan\u00eb zhdukur,” shkroi Crane.<\/p>\n\n\n\n
“Kjo nuk \u00ebsht\u00eb nj\u00eb histori p\u00ebr nj\u00eb agjent t\u00eb keq ose nj\u00eb API t\u00eb keq. B\u00ebhet fjal\u00eb p\u00ebr nj\u00eb industri t\u00eb t\u00ebr\u00eb q\u00eb nd\u00ebrton integrime IA-agjent\u00eb n\u00eb infrastruktur\u00ebn e prodhimit m\u00eb shpejt sesa po nd\u00ebrton arkitektur\u00ebn e siguris\u00eb p\u00ebr t’i b\u00ebr\u00eb k\u00ebto integrime t\u00eb sigurta,” shtoi ai.<\/p>\n\n\n\n
Crane m\u00eb von\u00eb konfirmoi t\u00eb h\u00ebn\u00ebn, dy dit\u00eb pas incidentit, se t\u00eb dh\u00ebnat e humbura ishin rikuperuar.<\/p>\n\n\n\n
Incidenti vjen nd\u00ebrsa modelet e IA-s\u00eb b\u00ebhen m\u00eb t\u00eb sofistikuara, ve\u00e7an\u00ebrisht q\u00eb nga njoftimi i modelit t\u00eb fundit t\u00eb Anthropic, Mythos, dhe bankier\u00ebt dhe qeverit\u00eb japin alarmin p\u00ebr incidente t\u00eb mundshme t\u00eb siguris\u00eb kibernetike.<\/p>\n","protected":false},"excerpt":{"rendered":"
Sistemi i inteligjenc\u00ebs artificiale, i mund\u00ebsuar nga modeli Claude Opus i Anthropic, kishte qen\u00eb duke trajtuar nj\u00eb detyr\u00eb rutin\u00eb kur zgjodhi n\u00eb m\u00ebnyr\u00eb t\u00eb pavarur t\u00eb “rregullonte” nj\u00eb problem duke fshir\u00eb t\u00eb dh\u00ebnat – pa asnj\u00eb miratim njer\u00ebzor. \u00c7far\u00eb tmerri! Nj\u00eb agjent i inteligjenc\u00ebs artificiale i projektuar p\u00ebr t\u00eb p\u00ebrmir\u00ebsuar detyrat e kodimit arriti t\u00eb<\/p>\n","protected":false},"author":4,"featured_media":26772,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[349],"tags":[40],"class_list":{"0":"post-26771","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-featured"},"yoast_head":"\n