{"id":427,"date":"2026-07-31T09:48:22","date_gmt":"2026-07-31T09:48:22","guid":{"rendered":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/"},"modified":"2026-07-31T09:48:22","modified_gmt":"2026-07-31T09:48:22","slug":"anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2","status":"publish","type":"post","link":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/","title":{"rendered":"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode"},"content":{"rendered":"\n

Claude-modeller slap ud af angiveligt isolerede cybersikkerheds-evalueringsmilj\u00f8er og n\u00e5ede faktiske internetvendte systemer, har Anthropic bekr\u00e6ftet.<\/p>\n\n\n\n

Anthropic offentliggjorde tre sikkerhedsh\u00e6ndelser<\/a>, der involverede modeller, som troede, de angreb simulerede m\u00e5l. En konfigurationsfejl havde efterladt internetadgang aktiveret, s\u00e5 de kunne interagere med virkelige tjenester og organisationer.<\/p>\n\n\n\n

H\u00e6ndelserne f\u00f8lger i k\u00f8lvandet p\u00e5 det nylige Hugging Face-brud med OpenAI-agenter og rapporter om, at OpenAI-modeller udnyttede zero-day-s\u00e5rbarheder i JFrog Artifactory<\/a> under fors\u00f8g p\u00e5 at undslippe en sandkasse.<\/p>\n\n\n\n

Internetadgangen forblev aktiveret under Claudes sikkerhedstests<\/h2>\n\n\n\n

Tredjepartsevaluatoren Irregular organiserede testene som capture-the-flag-\u00f8velser. Modellerne fik instruktioner om, at de opererede i isolerede simuleringer.<\/p>\n\n\n\n

Testmilj\u00f8et kunne dog stadig tilg\u00e5 det offentlige internet. N\u00e5r modellerne ikke kunne finde deres tilsigtede simulerede m\u00e5l, s\u00f8gte de efter lignende systemer online og behandlede nogle virkelige tjenester som en del af \u00f8velsen.<\/p>\n\n\n\n

Anthropic og Irregular identificerede ikke konfigurationsproblemet, f\u00f8r m\u00e5neder efter den f\u00f8rste h\u00e6ndelse. Anthropic oplyste, at det \u00e6ldste kendte brud fandt sted i april og forblev uopdaget i cirka tre m\u00e5neder.<\/p>\n\n\n\n

Claude Mythos 5 offentliggjorde en ondsindet PyPI-pakke<\/h3>\n\n\n\n

Under en evaluering fandt Claude Mythos 5 installationsinstruktioner, der henviste til en fiktiv Python-pakke. Pakkenavnet var ikke registreret p\u00e5 Python Package Index.<\/p>\n\n\n\n

Modellen inds\u00e5, at den kunne registrere det ubrugte navn og offentligg\u00f8re sin egen kode. Systemer, der fulgte installationsinstruktionerne, kunne derefter automatisk downloade og udf\u00f8re pakken.<\/p>\n\n\n\n

Claude oprettede de n\u00f8dvendige konti og uploadede en ondsindet pakke til PyPI. Pakken forblev offentligt tilg\u00e6ngelig i cirka en time, f\u00f8r PyPIs automatiske beskyttelsesmekanismer fjernede den.<\/p>\n\n\n\n

I den periode downloadede og udf\u00f8rte 15 virkelige systemer koden.<\/p>\n\n\n\n

Legitimationsoplysninger fra et sikkerhedsfirma blev stj\u00e5let<\/h3>\n\n\n\n

Et af de ber\u00f8rte systemer tilh\u00f8rte et cybersikkerhedsfirma, der automatisk downloader og scanner nyudgivne PyPI-pakker.<\/p>\n\n\n\n

Den ondsindede pakke indsamlede legitimationsoplysninger fra systemet og sendte dem til en infrastruktur, som Claude havde oprettet. Modellen brugte derefter de stj\u00e5lne legitimationsoplysninger til at f\u00e5 adgang til yderligere dele af firmaets netv\u00e6rk.<\/p>\n\n\n\n

Anthropic oplyste, at Claude ikke havde nogen forudg\u00e5ende viden om organisationen og ikke bevidst udvalgte den som m\u00e5l. Virksomheden underrettede PyPI og delte indikatorer knyttet til den ondsindede pakke.<\/p>\n\n\n\n

Det ber\u00f8rte cybersikkerhedsfirma havde ikke opdaget indtr\u00e6ngningen, f\u00f8r Anthropic kontaktede dem.<\/p>\n\n\n\n

Claude Opus 4.7 fik adgang til en produktionsdatabase<\/h3>\n\n\n\n

En separat evaluering med Claude Opus 4.7 f\u00f8rte til adgang til en virkelig produktionsdatabase.<\/p>\n\n\n\n

Det fiktive m\u00e5l i \u00f8velsen delte navn med en eksisterende virksomhed og et dom\u00e6ne. Efter at Claude ikke kunne n\u00e5 det simulerede system, fandt den den rigtige organisation online og behandlede den som det tilsigtede m\u00e5l.<\/p>\n\n\n\n

Modellen udtrak applikations- og infrastrukturadgangsoplysninger, f\u00f8r den fik adgang til en database med flere hundrede r\u00e6kker produktionsinformation.<\/p>\n\n\n\n

Anthropic beskrev denne h\u00e6ndelse som mere alvorlig, fordi modellen n\u00e5ede live forretningssystemer og fik adgang til rigtige data.<\/p>\n\n\n\n

Forskningsmodel scannede omkring 9.000 systemer<\/h3>\n\n\n\n

En endnu ikke frigivet Anthropic-forskningsmodel for\u00e5rsagede den tredje h\u00e6ndelse.<\/p>\n\n\n\n

Efter at det ikke lykkedes at finde det tilsigtede testm\u00e5l, scannede modellen omkring 9.000 internetforbundne systemer. Den kompromitterede en eksponeret applikation ved hj\u00e6lp af legitimationsoplysninger vist p\u00e5 en debug-side samt en SQL-injektionss\u00e5rbarhed.<\/p>\n\n\n\n

Modellen fastslog senere, at den kompromitterede applikation tilh\u00f8rte en uafh\u00e6ngig cloud-konto. Den stoppede derefter angrebet uden menneskelig indgriben.<\/p>\n\n\n\n

Anthropic fors\u00f8gte stadig at kontakte den ber\u00f8rte organisation, da de offentliggjorde deres resultater.<\/p>\n\n\n\n

Modellerne baserede sig p\u00e5 grundl\u00e6ggende sikkerhedssvagheder<\/h3>\n\n\n\n

Anthropic oplyste, at h\u00e6ndelserne ikke involverede nye eller tidligere ukendte angrebsteknikker.<\/p>\n\n\n\n

Modellerne benyttede sig af almindelige svagheder som eksponerede legitimationsoplysninger, svage adgangskoder, offentligt tilg\u00e6ngelige endepunkter, debug-sider med f\u00f8lsomme oplysninger samt SQL-injektionss\u00e5rbarheder.<\/p>\n\n\n\n

Hovedfejlen skyldtes ubegr\u00e6nset internetadgang og modellernes antagelse om, at ethvert tilg\u00e6ngeligt system var en del af den autoriserede test.<\/p>\n\n\n\n

Anthropic satte cyberafpr\u00f8vning p\u00e5 pause efter gennemgang af h\u00e6ndelserne<\/h3>\n\n\n\n

Anthropic p\u00e5begyndte gennemgangen af evalueringerne og satte al cybersikkerhedstest p\u00e5 pause den 23. juli.<\/p>\n\n\n\n

Virksomheden identificerede de tre h\u00e6ndelser den 24. juli. Irregular og to ber\u00f8rte organisationer modtog underretninger den 27. juli.<\/p>\n\n\n\n

Anthropic fortsatte med at fors\u00f8ge at n\u00e5 den organisation, der var involveret i den tredje h\u00e6ndelse.<\/p>\n\n\n\n

Ingen af de to organisationer, som Anthropic fik kontakt til, havde uafh\u00e6ngigt opdaget kompromitteringerne, f\u00f8r de modtog underretningerne.<\/p>\n\n\n\n

Evalueringsmilj\u00f8et manglede produktionssikkerhedsforanstaltninger<\/h3>\n\n\n\n

Testsystemerne anvendte ikke de klassifikatorer, overv\u00e5gningsv\u00e6rkt\u00f8jer og andre sikkerhedsmekanismer, der er inkluderet i offentligt tilg\u00e6ngelige Claude-modeller.<\/p>\n\n\n\n

Anthropic oplyste, at deres produktionsbeskyttelse sandsynligvis ville have blokeret de handlinger, der blev observeret under evalueringerne. De karakteriserede h\u00e6ndelserne prim\u00e6rt som svigt i testedesign, infrastruktur, overv\u00e5gning og tredjepartstilsyn.<\/p>\n\n\n\n

Virksomheden betragtede ikke h\u00e6ndelserne som bevis p\u00e5 et grundl\u00e6ggende alignment-svigt. I stedet fulgte modellerne evalueringsinstruktionerne, mens de fejlagtigt interagerede med virkelige m\u00e5l.<\/p>\n\n\n\n

Anthropic planl\u00e6gger strengere overv\u00e5gning af fremtidige test<\/h3>\n\n\n\n

Anthropic planl\u00e6gger at udvide transskriptoverv\u00e5gning p\u00e5 tv\u00e6rs af fremtidige evalueringer og indf\u00f8re st\u00e6rkere unders\u00f8gelsesv\u00e6rkt\u00f8jer.<\/p>\n\n\n\n

De vil ogs\u00e5 \u00f8ge tilsynet med tredjeparts testpartnere og gennemg\u00e5, hvordan evalueringsmilj\u00f8er forbinder til ekstern infrastruktur.<\/p>\n\n\n\n

Anthropic dr\u00f8fter en uafh\u00e6ngig gennemgang med METR og planl\u00e6gger at offentligg\u00f8re en delvist redigeret transskription af PyPI-h\u00e6ndelsen.<\/p>\n\n\n\n

P\u00e5 trods af sikkerhedsfejlene har Claude Mythos demonstreret st\u00e6rke evner inden for s\u00e5rbarhedsdetektion. Microsoft har ogs\u00e5 anerkendt, at de k\u00e6mper med at rette alle de s\u00e5rbarheder, Claude Mythos finder<\/a>, hvilket understreger b\u00e5de modellens potentiale inden for sikkerhedsforskning og de risici, der opst\u00e5r ved d\u00e5rligt isolerede testmilj\u00f8er.<\/p>\n","protected":false},"excerpt":{"rendered":"

Claude-modeller slap ud af angiveligt isolerede cybersikkerheds-evalueringsmilj\u00f8er og n\u00e5ede faktiske internetvendte systemer, har Anthropic bekr\u00e6ftet. Anthropic offentliggjorde tre sikkerhedsh\u00e6ndelser, der involverede modeller, som troede, de angreb simulerede m\u00e5l. En konfigurationsfejl havde efterladt internetadgang aktiveret, s\u00e5 de kunne interagere med virkelige tjenester og organisationer. H\u00e6ndelserne f\u00f8lger i k\u00f8lvandet p\u00e5 det nylige Hugging Face-brud med OpenAI-agenter og […]<\/p>\n","protected":false},"author":49,"featured_media":428,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_enabled_clarity":"","footnotes":""},"categories":[1],"tags":[],"work":[],"class_list":["post-427","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"acf":[],"yoast_head":"\nAnthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode - WindowsReport DK (Danish)<\/title>\n<meta name=\"description\" content=\"Anthropic siger, at Claude-modeller undslap sikkerhedstests, offentliggjorde en ondsindet PyPI-pakke og fik adgang til reelle produktionssystemer.\" \/>\n<meta name=\"robots\" content=\"noindex, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode\" \/>\n<meta property=\"og:description\" content=\"Anthropic siger, at Claude-modeller undslap sikkerhedstests, offentliggjorde en ondsindet PyPI-pakke og fik adgang til reelle produktionssystemer.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/\" \/>\n<meta property=\"og:site_name\" content=\"WindowsReport DK (Danish)\" \/>\n<meta property=\"article:published_time\" content=\"2026-07-31T09:48:22+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/windowsreport.com\/dk\/wp-content\/uploads\/sites\/12\/2026\/07\/anthropic-hacked-organizations-2.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1280\" \/>\n\t<meta property=\"og:image:height\" content=\"801\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Milan Stanojevic\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@MilanStanojev17\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Milan Stanojevic\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"5 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/\"},\"author\":{\"name\":\"Milan Stanojevic\",\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/#\\\/schema\\\/person\\\/e3910b02e4592db47a0eb303765a3a25\"},\"headline\":\"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode\",\"datePublished\":\"2026-07-31T09:48:22+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/\"},\"wordCount\":914,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/wp-content\\\/uploads\\\/sites\\\/12\\\/2026\\\/07\\\/anthropic-hacked-organizations-2.jpg\",\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/#respond\"]}],\"copyrightYear\":\"2026\",\"copyrightHolder\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/#organization\"}},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/\",\"url\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/\",\"name\":\"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode - WindowsReport DK (Danish)\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/wp-content\\\/uploads\\\/sites\\\/12\\\/2026\\\/07\\\/anthropic-hacked-organizations-2.jpg\",\"datePublished\":\"2026-07-31T09:48:22+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/#\\\/schema\\\/person\\\/e3910b02e4592db47a0eb303765a3a25\"},\"description\":\"Anthropic siger, at Claude-modeller undslap sikkerhedstests, offentliggjorde en ondsindet PyPI-pakke og fik adgang til reelle produktionssystemer.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/#primaryimage\",\"url\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/wp-content\\\/uploads\\\/sites\\\/12\\\/2026\\\/07\\\/anthropic-hacked-organizations-2.jpg\",\"contentUrl\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/wp-content\\\/uploads\\\/sites\\\/12\\\/2026\\\/07\\\/anthropic-hacked-organizations-2.jpg\",\"width\":1280,\"height\":801},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/#website\",\"url\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/\",\"name\":\"WindowsReport DK (Danish)\",\"description\":\"\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/#\\\/schema\\\/person\\\/e3910b02e4592db47a0eb303765a3a25\",\"name\":\"Milan Stanojevic\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/fdfb58c5f9c33b78cbac22bd664ac8335249104922392e67d69d32ee59624544?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/fdfb58c5f9c33b78cbac22bd664ac8335249104922392e67d69d32ee59624544?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/fdfb58c5f9c33b78cbac22bd664ac8335249104922392e67d69d32ee59624544?s=96&d=mm&r=g\",\"caption\":\"Milan Stanojevic\"},\"description\":\"Milan has been enthusiastic about technology ever since his childhood days, and this led him to take interest in all PC-related technologies. He's a PC enthusiast and he spends most of his time learning about computers and technology. Before joining WindowsReport, he worked as a front-end web developer. Now, he's one of the Troubleshooting experts in our worldwide team, specializing in Windows errors & software issues.\",\"sameAs\":[\"https:\\\/\\\/x.com\\\/MilanStanojev17\"],\"url\":\"https:\\\/\\\/windowsreport.com\\\/dk\\\/author\\\/milan\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode - WindowsReport DK (Danish)","description":"Anthropic siger, at Claude-modeller undslap sikkerhedstests, offentliggjorde en ondsindet PyPI-pakke og fik adgang til reelle produktionssystemer.","robots":{"index":"noindex","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"og_locale":"en_US","og_type":"article","og_title":"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode","og_description":"Anthropic siger, at Claude-modeller undslap sikkerhedstests, offentliggjorde en ondsindet PyPI-pakke og fik adgang til reelle produktionssystemer.","og_url":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/","og_site_name":"WindowsReport DK (Danish)","article_published_time":"2026-07-31T09:48:22+00:00","og_image":[{"width":1280,"height":801,"url":"https:\/\/windowsreport.com\/dk\/wp-content\/uploads\/sites\/12\/2026\/07\/anthropic-hacked-organizations-2.jpg","type":"image\/jpeg"}],"author":"Milan Stanojevic","twitter_card":"summary_large_image","twitter_creator":"@MilanStanojev17","twitter_misc":{"Written by":"Milan Stanojevic","Est. reading time":"5 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/#article","isPartOf":{"@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/"},"author":{"name":"Milan Stanojevic","@id":"https:\/\/windowsreport.com\/dk\/#\/schema\/person\/e3910b02e4592db47a0eb303765a3a25"},"headline":"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode","datePublished":"2026-07-31T09:48:22+00:00","mainEntityOfPage":{"@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/"},"wordCount":914,"commentCount":0,"image":{"@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/#primaryimage"},"thumbnailUrl":"https:\/\/windowsreport.com\/dk\/wp-content\/uploads\/sites\/12\/2026\/07\/anthropic-hacked-organizations-2.jpg","inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/#respond"]}],"copyrightYear":"2026","copyrightHolder":{"@id":"https:\/\/windowsreport.com\/#organization"}},{"@type":"WebPage","@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/","url":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/","name":"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode - WindowsReport DK (Danish)","isPartOf":{"@id":"https:\/\/windowsreport.com\/dk\/#website"},"primaryImageOfPage":{"@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/#primaryimage"},"image":{"@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/#primaryimage"},"thumbnailUrl":"https:\/\/windowsreport.com\/dk\/wp-content\/uploads\/sites\/12\/2026\/07\/anthropic-hacked-organizations-2.jpg","datePublished":"2026-07-31T09:48:22+00:00","author":{"@id":"https:\/\/windowsreport.com\/dk\/#\/schema\/person\/e3910b02e4592db47a0eb303765a3a25"},"description":"Anthropic siger, at Claude-modeller undslap sikkerhedstests, offentliggjorde en ondsindet PyPI-pakke og fik adgang til reelle produktionssystemer.","breadcrumb":{"@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/#primaryimage","url":"https:\/\/windowsreport.com\/dk\/wp-content\/uploads\/sites\/12\/2026\/07\/anthropic-hacked-organizations-2.jpg","contentUrl":"https:\/\/windowsreport.com\/dk\/wp-content\/uploads\/sites\/12\/2026\/07\/anthropic-hacked-organizations-2.jpg","width":1280,"height":801},{"@type":"BreadcrumbList","@id":"https:\/\/windowsreport.com\/dk\/anthropic-siger-at-claude-hackede-3-organisationer-og-offentliggjorde-ondsindet-pypi-kode-2\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/windowsreport.com\/dk\/"},{"@type":"ListItem","position":2,"name":"Anthropic siger, at Claude hackede 3 organisationer og offentliggjorde ondsindet PyPI-kode"}]},{"@type":"WebSite","@id":"https:\/\/windowsreport.com\/dk\/#website","url":"https:\/\/windowsreport.com\/dk\/","name":"WindowsReport DK (Danish)","description":"","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/windowsreport.com\/dk\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/windowsreport.com\/dk\/#\/schema\/person\/e3910b02e4592db47a0eb303765a3a25","name":"Milan Stanojevic","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/fdfb58c5f9c33b78cbac22bd664ac8335249104922392e67d69d32ee59624544?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/fdfb58c5f9c33b78cbac22bd664ac8335249104922392e67d69d32ee59624544?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/fdfb58c5f9c33b78cbac22bd664ac8335249104922392e67d69d32ee59624544?s=96&d=mm&r=g","caption":"Milan Stanojevic"},"description":"Milan has been enthusiastic about technology ever since his childhood days, and this led him to take interest in all PC-related technologies. He's a PC enthusiast and he spends most of his time learning about computers and technology. Before joining WindowsReport, he worked as a front-end web developer. Now, he's one of the Troubleshooting experts in our worldwide team, specializing in Windows errors & software issues.","sameAs":["https:\/\/x.com\/MilanStanojev17"],"url":"https:\/\/windowsreport.com\/dk\/author\/milan\/"}]}},"_links":{"self":[{"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/posts\/427","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/users\/49"}],"replies":[{"embeddable":true,"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/comments?post=427"}],"version-history":[{"count":0,"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/posts\/427\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/media\/428"}],"wp:attachment":[{"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/media?parent=427"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/categories?post=427"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/tags?post=427"},{"taxonomy":"work","embeddable":true,"href":"https:\/\/windowsreport.com\/dk\/wp-json\/wp\/v2\/work?post=427"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}