{"id":168672,"date":"2026-04-14T11:12:44","date_gmt":"2026-04-14T09:12:48","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/v-linux-utverdili-oficzialnuyu-politiku-ispolzovaniya-ii-koda-posle-mesyaczev-ostryh-sporov"},"modified":"2026-04-16T10:52:23","modified_gmt":"2026-04-16T08:52:23","slug":"v-linux-utverdili-oficzialnuyu-politiku-ispolzovaniya-ii-koda-posle-mesyaczev-ostryh-sporov","status":"publish","type":"post","link":"https:\/\/prohoster.info\/cs\/blog\/news\/v-linux-utverdili-oficzialnuyu-politiku-ispolzovaniya-ii-koda-posle-mesyaczev-ostryh-sporov","title":{"rendered":"V Linuxu byla schv\u00e1lena ofici\u00e1ln\u00ed politika pou\u017e\u00edv\u00e1n\u00ed AI k\u00f3du po m\u011bs\u00edc\u00edch ostr\u00fdch debat","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

V\u00fdvoj\u00e1\u0159i j\u00e1dra Linux ofici\u00e1ln\u011b schv\u00e1lili politiku pou\u017e\u00edv\u00e1n\u00ed um\u011bl\u00e9 inteligence (AI) po m\u011bs\u00edc\u00edch bou\u0159liv\u00fdch debat. Linus Torvalds a pr\u016fmyslov\u00ed partne\u0159i dosp\u011bli k dohod\u011b, kter\u00e1 povoluje pou\u017e\u00edv\u00e1n\u00ed AI-asistent\u016f, ale zav\u00e1d\u00ed p\u0159\u00edsn\u00e1 pravidla pro zve\u0159ejn\u011bn\u00ed informac\u00ed a klade plnou odpov\u011bdnost za jak\u00e9koli chyby na lidsk\u00e9 v\u00fdvoj\u00e1\u0159e.<\/p>\n

\"V<\/p>\n

Nov\u00e1 pravidla zakazuj\u00ed pou\u017e\u00edv\u00e1n\u00ed pr\u00e1vn\u011b z\u00e1vazn\u00e9ho tagu \u201eSigned-off-by\u201c pro k\u00f3d vytvo\u0159en\u00fd AI syst\u00e9my, nahrazuj\u00ed jej nov\u00fdm tagem \u201eAssisted-by\u201c pro zaji\u0161t\u011bn\u00ed transparentnosti p\u016fvodu \u00faprav. Takov\u00e9 rozhodnut\u00ed, jak uv\u00e1d\u00ed Tom’s Hardware<\/a>, reagovalo na rostouc\u00ed nap\u011bt\u00ed v komunit\u011b zp\u016fsoben\u00e9 p\u0159\u00edlivem nekvalitn\u00edho k\u00f3du a skryt\u00fdm pou\u017e\u00edv\u00e1n\u00edm neuronov\u00fdch s\u00edt\u00ed. Nyn\u00ed bude ka\u017ed\u00e1 chyba nebo zranitelnost v takov\u00e9m k\u00f3du pr\u00e1vn\u011b p\u0159id\u011blena konkr\u00e9tn\u00ed osob\u011b, kter\u00e1 publikovala zm\u011bny. To vylou\u010d\u00ed mo\u017enost p\u0159en\u00e1\u0161en\u00ed viny na algoritmy a umo\u017en\u00ed integrovat modern\u00ed n\u00e1stroje do pracovn\u00edho procesu, ani\u017e by to poru\u0161ovalo principy certifikace Developer Certificate of Origin (DCO).<\/p>\n

\"V<\/p>\n

Neshody kolem integrace AI dos\u00e1hly vrcholu v lednu, kdy se z\u00e1stupci spole\u010dnosti Intel a Oracle ve\u0159ejn\u011b dostali do konfliktu ohledn\u011b regulace neuronov\u00fdch s\u00edt\u00ed v projektu. P\u0159i\u010dem\u017e Torvalds ost\u0159e kritizoval my\u0161lenky jejich \u00fapln\u00e9ho z\u00e1kazu, nazval to bezv\u00fdznamn\u00fdm, a trval na tom, \u017ee AI je prost\u011b n\u00e1stroj, jako ka\u017ed\u00fd jin\u00fd software. Zd\u016fraznil, \u017ee zlo\u010dinci stejn\u011b nebudou \u010d\u00edst dokumentaci, tak\u017ee je efektivn\u011bj\u0161\u00ed kontrolovat v\u00fdsledek prost\u0159ednictv\u00edm odpov\u011bdnosti konkr\u00e9tn\u00edch lid\u00ed, ne\u017e se sna\u017eit zak\u00e1zat technologie na lok\u00e1ln\u00edch stroj\u00edch v\u00fdvoj\u00e1\u0159\u016f.<\/p>\n

P\u0159ed p\u0159ijet\u00edm jednotn\u00e9 politiky m\u011bly velk\u00e9 projekty otev\u0159en\u00e9ho softwaru r\u016fzn\u00e9 p\u0159\u00edstupy k probl\u00e9mu, a to a\u017e po \u00fapln\u00e9 z\u00e1kazy. Nap\u0159\u00edklad distribuce Gentoo a NetBSD d\u0159\u00edve zablokovaly p\u0159\u00edjem generovan\u00e9ho k\u00f3du AI, argumentuj\u00edc t\u00edm, \u017ee data pro tr\u00e9nink model\u016f \u010dasto obsahuj\u00ed k\u00f3d s restriktivn\u00edmi licencemi, co\u017e \u010din\u00ed v\u00fdsledek pr\u00e1vn\u011b neplatn\u00fdm. Odborn\u00edci z Red Hat varovali, \u017ee pou\u017e\u00edv\u00e1n\u00ed t\u011bchto n\u00e1stroj\u016f m\u016f\u017ee nev\u011bdomky poru\u0161ovat licence GNU GPL a podkop\u00e1vat cel\u00fd syst\u00e9m DCO, jeliko\u017e v\u00fdvoj\u00e1\u0159 nem\u016f\u017ee zaru\u010dit \u010distotu p\u016fvodu ka\u017ed\u00e9ho \u0159\u00e1dku k\u00f3du.<\/p>\n

Krom\u011b pr\u00e1vn\u00edch rizik se komunita setkala s probl\u00e9mem obrovsk\u00e9ho objemu nekvalitn\u00edch \u00faprav, kter\u00e9 se naz\u00fdvaj\u00ed AI slop. Tv\u016frci popul\u00e1rn\u00edch projekt\u016f, jako jsou cURL a tldraw, byli nuceni uzav\u0159\u00edt programy odm\u011bn nebo automaticky odm\u00edtat extern\u00ed \u017e\u00e1dosti kv\u016fli spamu obsahuj\u00edc\u00edmu k\u00f3d s AI halucinacemi. V p\u0159\u00edpad\u011b j\u00e1dra Linuxu se situace zhor\u0161ila po skand\u00e1ln\u00edm incidentu s maintainerem Saschou Levinem, in\u017een\u00fdrem z Nvidia, kter\u00fd implementoval patch pro j\u00e1dro 6.15, pln\u011b napsan\u00fd modelem AI, bez odpov\u00eddaj\u00edc\u00edho upozorn\u011bn\u00ed. K\u00f3d byl funk\u010dn\u00ed, av\u0161ak obsahoval regresi v\u00fdkonu, p\u0159esto\u017ee byl zkontrolov\u00e1n a testov\u00e1n.<\/p>\n

Probl\u00e9m ned\u016fv\u011bry v d\u016fsledku skryt\u00e9ho pou\u017e\u00edv\u00e1n\u00ed AI se dotkl nejen syst\u00e9mov\u00e9ho programov\u00e1n\u00ed, ale i dal\u0161\u00edch oblast\u00ed, nap\u0159\u00edklad komunity modder\u016f Doom. Konkr\u00e9tn\u011b se l\u00eddr projektu GZDoom Christoph Oelckers setkal s masov\u00fdm odlivem \u00fa\u010dastn\u00edk\u016f projektu pot\u00e9, co pou\u017eil AI patche bez odhalen\u00ed t\u00e9to skute\u010dnosti a odm\u00edtl ud\u011blat kompromis s kritiky. V d\u016fsledku toho v\u011bt\u0161ina p\u0159isp\u011bvatel\u016f opustila p\u016fvodn\u00ed projekt a vytvo\u0159ila fork UZDoom, co\u017e uk\u00e1zalo d\u016fle\u017eitost otev\u0159enosti a up\u0159\u00edmnosti p\u0159ed komunitou.<\/p>\n

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