{"id":359810,"date":"2026-09-25T16:13:37","date_gmt":"2026-09-25T14:13:37","guid":{"rendered":"https:\/\/legrandcontinent.eu\/fr\/?p=359810"},"modified":"2026-09-25T19:22:14","modified_gmt":"2026-09-25T17:22:14","slug":"washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante","status":"publish","type":"post","link":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/","title":{"rendered":"Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les \u00e0 une \u00e9valuation ind\u00e9pendante"},"content":{"rendered":"\n<p>La Maison-Blanche a demand\u00e9 \u00e0 certains laboratoires am\u00e9ricains d\u2019IA, dont OpenAI et Anthropic, de r\u00e9server l\u2019acc\u00e8s \u00e0 leurs nouveaux mod\u00e8les de fronti\u00e8re aux tests du gouvernement am\u00e9ricain avant de les partager avec d\u2019autres organismes d\u2019\u00e9valuation ind\u00e9pendants, comme l\u2019AI Security Institute (UK AISI), qui fait r\u00e9f\u00e9rence dans le domaine&nbsp;<span class='whitespace-nowrap'><span id='easy-footnote-1-359810' class='easy-footnote-margin-adjust'><\/span><span class='easy-footnote note-digits-1'><a href='https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#easy-footnote-bottom-1-359810' title='Sophia Cai et Joseph Bambridge, \u00ab&amp;#160;&lt;a href=&quot;https:\/\/www.politico.com\/news\/2026\/09\/24\/white-house-asks-openai-and-anthropic-to-hold-new-models-from-uk-testers-until-u-s-review-01091769&quot;&gt;White House asks OpenAI and Anthropic to hold new models from UK testers until US review&lt;\/a&gt;&amp;#160;\u00bb, &lt;em&gt;Politico&lt;\/em&gt;, 24 septembre 2026.'><sup>1<\/sup><\/a><\/span><\/span>.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>L\u2019administration am\u00e9ricaine remettrait ainsi en cause l\u2019acc\u00e8s privil\u00e9gi\u00e9 pr\u00e9-d\u00e9ploiement dont disposait jusqu\u2019ici cet institut public britannique.<\/li>\n\n\n\n<li>L\u2019ordre d\u2019acc\u00e8s aux nouveaux mod\u00e8les de fronti\u00e8re serait d\u00e9sormais le suivant&#160;: d\u2019abord, une \u00e9valuation par le gouvernement am\u00e9ricain, puis la s\u00e9curisation des syst\u00e8mes am\u00e9ricains, et seulement ensuite le partage du mod\u00e8le avec des partenaires \u00e9trangers.<\/li>\n<\/ul>\n\n\n\n<p>L&rsquo;institut UK AISI est l\u2019un des principaux organismes publics au monde sp\u00e9cialis\u00e9s dans l\u2019\u00e9valuation pr\u00e9 et post-d\u00e9ploiement des mod\u00e8les de fronti\u00e8re.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Cr\u00e9\u00e9 en novembre 2023 \u00e0 la suite du sommet sur la s\u00e9curit\u00e9 de l\u2019IA de Bletchley Park, il \u00e9value la plupart des mod\u00e8les \u00e0 la fronti\u00e8re et documente notamment leurs capacit\u00e9s cyber, biologiques et chimiques, l\u2019autonomie des agents ou encore les risques li\u00e9s \u00e0 l\u2019auto-am\u00e9lioration r\u00e9cursive de l\u2019IA.<\/li>\n\n\n\n<li>Comme d\u2019autres organisations ind\u00e9pendantes d\u2019\u00e9valuation (METR, Redwood Research notamment), le UK AISI pouvait b\u00e9n\u00e9ficier d\u2019un acc\u00e8s anticip\u00e9 \u00e0 des checkpoints pr\u00e9-d\u00e9ploiement grand public afin d\u2019effectuer les tests sur ses propres benchmarks pour documenter la progression des capacit\u00e9s et des risques associ\u00e9s aux mod\u00e8les d\u2019IA.<\/li>\n<\/ul>\n\n\n\n<div class=\"iframe-container wp-block-image wp-block-image-medium  iframe-dw\">\n\t<div>\n\t\t<iframe class=\"absolute w-full h-full pin-t pin-l\" title=\"titolo\" aria-label=\"Interactive line chart\" src=\"\/\/datawrapper.dwcdn.net\/rA8m8\" scrolling=\"no\" frameborder=\"0\"><\/iframe>\n\t<\/div>\n<\/div>\n\n\n\n<p>D\u00e9but septembre, le <em>Financial Times<\/em> rapportait que le UK AISI n\u2019avait pas eu acc\u00e8s \u00e0 Mythos 5.1, alors qu\u2019il disposait historiquement d\u2019un acc\u00e8s anticip\u00e9 aux principaux mod\u00e8les am\u00e9ricains&nbsp;<span class='whitespace-nowrap'><span id='easy-footnote-2-359810' class='easy-footnote-margin-adjust'><\/span><span class='easy-footnote note-digits-1'><a href='https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#easy-footnote-bottom-2-359810' title='Lucy Fisher et Madhumita Murgia, \u00ab&amp;#160;&lt;a href=&quot;https:\/\/www.ft.com\/content\/560e1c8b-f163-4fd6-b604-e905550ac870?syn-25a6b1a6=1&quot;&gt;Anthropic withheld latest AI model from UK testing agency&lt;\/a&gt;&amp;#160;\u00bb, &lt;em&gt;Financial Times&lt;\/em&gt;, 9 septembre 2026.'><sup>2<\/sup><\/a><\/span><\/span>. Auparavant, OpenAI, Anthropic et Google DeepMind ont tous fourni \u00e0 l\u2019Institut un acc\u00e8s \u00e0 certains de leurs mod\u00e8les les plus avanc\u00e9s afin qu\u2019ils soient test\u00e9s.<\/p>\n\n\n\n<p>Cette \u00e9volution intervient alors que les capacit\u00e9s cyber des mod\u00e8les de fronti\u00e8re deviennent un sujet de <a href=\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/la-chine-et-les-etats-unis-peuvent-ils-saccorder-sur-la-securite-de-lia\/\">s\u00e9curit\u00e9<\/a> nationale.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>En juin, l\u2019administration am\u00e9ricaine avait d\u00e9j\u00e0 impos\u00e9 des <a href=\"https:\/\/legrandcontinent.eu\/fr\/2026\/06\/13\/la-maison-blanche-force-anthropic-a-couper-ses-derniers-modeles-aux-etrangers\/\">contr\u00f4les<\/a> \u00e0 l\u2019exportation sur Fable et Mythos, les mod\u00e8les d\u2019Anthropic, notamment en raison de l\u2019am\u00e9lioration rapide de leurs capacit\u00e9s dans le domaine de la cybers\u00e9curit\u00e9, interdisant ainsi leur utilisation par des ressortissants \u00e9trangers.<\/li>\n\n\n\n<li>La NSA teste d\u00e9sormais les mod\u00e8les avanc\u00e9s pour identifier des vuln\u00e9rabilit\u00e9s susceptibles d\u2019affecter les syst\u00e8mes de renseignement et de d\u00e9fense et d\u00e9veloppe ses propres \u00e9valuations. L\u2019agence y consacrerait plusieurs milliards de dollars par an, principalement en capacit\u00e9s de calcul&nbsp;<span class='whitespace-nowrap'><span id='easy-footnote-3-359810' class='easy-footnote-margin-adjust'><\/span><span class='easy-footnote note-digits-1'><a href='https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#easy-footnote-bottom-3-359810' title='Jeff Stein, \u00ab&amp;#160;&lt;a href=&quot;https:\/\/www.washingtonsun.com\/technology\/classified-estimates-nsa-paying-billions-to-test-ai-models&quot;&gt;Classified Estimates Show the NSA Is Paying Billions to Test AI Models&lt;\/a&gt;&amp;#160;\u00bb, &lt;em&gt;The Washington Sun&lt;\/em&gt;, 24 septembre 2026.'><sup>3<\/sup><\/a><\/span><\/span>.<\/li>\n\n\n\n<li>L\u2019\u00e9quivalent direct am\u00e9ricain du UK AISI, le Center for AI Standards and Innovation (CAISI), reste beaucoup plus limit\u00e9 en moyens que l\u2019institut britannique. H\u00e9berg\u00e9 au NIST, le CAISI est sous-financ\u00e9 et dispose d\u2019effectifs de recherche plus r\u00e9duits.<\/li>\n<\/ul>\n\n\n\n<p>Les laboratoires d\u2019IA ont \u00e9galement propos\u00e9 une \u00e9valuation directement entre laboratoires, sans supervision gouvernementale. Ils testeraient ainsi r\u00e9ciproquement leurs mod\u00e8les avant leur mise sur le march\u00e9.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Google, OpenAI et Anthropic travaillent conjointement \u00e0 la cr\u00e9ation d\u2019un organisme priv\u00e9 de normalisation consacr\u00e9 \u00e0 la s\u00e9curit\u00e9 de l\u2019IA, qui pourrait d\u00e9finir des standards communs d\u2019\u00e9valuation pr\u00e9-d\u00e9ploiement.<\/li>\n\n\n\n<li>La question autour des capacit\u00e9s d\u2019\u00e9valuation ind\u00e9pendantes devient d\u2019autant plus centrale que les capacit\u00e9s et les architectures des mod\u00e8les \u00e9voluent.<\/li>\n\n\n\n<li>Redwood Research souligne notamment que certaines architectures pourraient r\u00e9duire fortement la capacit\u00e9 d\u2019\u00e9valuateurs humains \u00e0 surveiller et auditer le comportement des mod\u00e8les, alors qu\u2019ils raisonnent non plus via une cha\u00eene de pens\u00e9e textuelle compr\u00e9hensible mais via des \u00e9tats latents internes non observables directement&nbsp;<span class='whitespace-nowrap'><span id='easy-footnote-4-359810' class='easy-footnote-margin-adjust'><\/span><span class='easy-footnote note-digits-1'><a href='https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#easy-footnote-bottom-4-359810' title='Lukas Finnveden &lt;em&gt;et. al.&lt;\/em&gt;, &lt;a href=&quot;https:\/\/www.redwoodresearch.org\/blog\/latent-reasoning-architectures-would&quot;&gt;Latent reasoning architectures would undermine CoT, our strongest oversight tool&lt;\/a&gt;, Redwood Research, 23 septembre 2026.'><sup>4<\/sup><\/a><\/span><\/span>.<\/li>\n<\/ul>\n\n\n\n<p>Au-del\u00e0 des tests et rapports d\u2019\u00e9valuation, cette asym\u00e9trie d\u2019acc\u00e8s aux mod\u00e8les fronti\u00e8res touche dans les usages l\u2019industrie europ\u00e9enne. Le projet Glasswing, le programme d\u2019Anthropic donnant un acc\u00e8s anticip\u00e9 \u00e0 Mythos, a permis d\u00e8s avril aux banques et entreprises critiques am\u00e9ricaines de renforcer leur d\u00e9fense cyber, notamment en recherchant avec Mythos des vuln\u00e9rabilit\u00e9s dans leurs propres syst\u00e8mes. Les banques europ\u00e9ennes n\u2019y ont eu acc\u00e8s que plusieurs mois plus tard.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Le gouvernement am\u00e9ricain serait le premier \u00e0 avoir acc\u00e8s aux nouveaux mod\u00e8les de fronti\u00e8re<\/p>\n","protected":false},"author":2622,"featured_media":359814,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"templates\/post-briefings.php","format":"standard","meta":{"_acf_changed":true,"_trash_the_other_posts":false,"_yoast_wpseo_estimated-reading-time-minutes":4,"_lgc_tts_voice":"","_lgc_tts_publish":false,"_lgc_tts_history_item_id":"","_lgc_tts_generated_at":"","_lgc_tts_status":"","_lgc_tts_error":"","footnotes":""},"categories":[3412],"tags":[],"staff":[5175],"editorial_format":[4942],"serie":[],"audience":[],"geo":[525],"class_list":["post-359810","post","type-post","status-publish","format-standard","hentry","category-puissances-de-lia","staff-henri-taurel","editorial_format-actu-breves","geo-ameriques"],"acf":{"open_in_webview":false,"accent":"default","_thumbnail_id":359814,"excerpt":"Le gouvernement am\u00e9ricain serait le premier \u00e0 avoir acc\u00e8s aux nouveaux mod\u00e8les de fronti\u00e8re","display_date":"","new_abstract":true},"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.1.1 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les d&#039;IA \u00e0 une \u00e9valuation ind\u00e9pendante<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les d&#039;IA \u00e0 une \u00e9valuation ind\u00e9pendante\" \/>\n<meta property=\"og:description\" content=\"Le gouvernement am\u00e9ricain serait le premier \u00e0 avoir acc\u00e8s aux nouveaux mod\u00e8les de fronti\u00e8re\" \/>\n<meta property=\"og:url\" content=\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/\" \/>\n<meta property=\"og:site_name\" content=\"Le Grand Continent\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-25T14:13:37+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-09-25T17:22:14+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/share-Washington-refuse-quAnthropic-et-OpenAI-soumettent-leurs-derniers-modeles-a-une-evaluation-independante.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1629\" \/>\n\t<meta property=\"og:image:height\" content=\"912\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Marin Saillofest\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:image\" content=\"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/share-Washington-refuse-quAnthropic-et-OpenAI-soumettent-leurs-derniers-modeles-a-une-evaluation-independante.png\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"Marin Saillofest\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/\",\"url\":\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/\",\"name\":\"Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les d'IA \u00e0 une \u00e9valuation ind\u00e9pendante\",\"isPartOf\":{\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/SIPA_01300557_000011-scaled.jpg\",\"datePublished\":\"2026-09-25T14:13:37+00:00\",\"dateModified\":\"2026-09-25T17:22:14+00:00\",\"author\":{\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/#\/schema\/person\/e4f85f52220fc7b4cc3cfdb86d88e833\"},\"breadcrumb\":{\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#breadcrumb\"},\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#primaryimage\",\"url\":\"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/SIPA_01300557_000011-scaled.jpg\",\"contentUrl\":\"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/SIPA_01300557_000011-scaled.jpg\",\"width\":2560,\"height\":1706,\"caption\":\"Sam Altman, Chief Executive Officer, OpenAI speaks during Security Council meeting on Artificial intelligence and international security at UN Headquarters in New York, NY on September 23, 2026. (Photo by Lev Radin\/Pacific Press) - LevRadin_092426(2) \/\/PACIFICPRESS_LevRadin_092426(2)\/Credit:Lev Radin\/PACIFIC PRESS\/SIPA\/2609241638\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Accueil\",\"item\":\"https:\/\/legrandcontinent.eu\/fr\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les \u00e0 une \u00e9valuation ind\u00e9pendante\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/#website\",\"url\":\"https:\/\/legrandcontinent.eu\/fr\/\",\"name\":\"Le Grand Continent\",\"description\":\"L&#039;\u00e9chelle pertinente\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/legrandcontinent.eu\/fr\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/#\/schema\/person\/e4f85f52220fc7b4cc3cfdb86d88e833\",\"name\":\"Marin Saillofest\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/legrandcontinent.eu\/fr\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/secure.gravatar.com\/avatar\/87088698b4d967390c49ae6d4f1aa459?s=96&d=mm&r=g\",\"contentUrl\":\"https:\/\/secure.gravatar.com\/avatar\/87088698b4d967390c49ae6d4f1aa459?s=96&d=mm&r=g\",\"caption\":\"Marin Saillofest\"}}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les d'IA \u00e0 une \u00e9valuation ind\u00e9pendante","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/","og_locale":"fr_FR","og_type":"article","og_title":"Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les d'IA \u00e0 une \u00e9valuation ind\u00e9pendante","og_description":"Le gouvernement am\u00e9ricain serait le premier \u00e0 avoir acc\u00e8s aux nouveaux mod\u00e8les de fronti\u00e8re","og_url":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/","og_site_name":"Le Grand Continent","article_published_time":"2026-09-25T14:13:37+00:00","article_modified_time":"2026-09-25T17:22:14+00:00","og_image":[{"width":1629,"height":912,"url":"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/share-Washington-refuse-quAnthropic-et-OpenAI-soumettent-leurs-derniers-modeles-a-une-evaluation-independante.png","type":"image\/png"}],"author":"Marin Saillofest","twitter_card":"summary_large_image","twitter_image":"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/share-Washington-refuse-quAnthropic-et-OpenAI-soumettent-leurs-derniers-modeles-a-une-evaluation-independante.png","twitter_misc":{"\u00c9crit par":"Marin Saillofest","Dur\u00e9e de lecture estim\u00e9e":"4 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/","url":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/","name":"Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les d'IA \u00e0 une \u00e9valuation ind\u00e9pendante","isPartOf":{"@id":"https:\/\/legrandcontinent.eu\/fr\/#website"},"primaryImageOfPage":{"@id":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#primaryimage"},"image":{"@id":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#primaryimage"},"thumbnailUrl":"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/SIPA_01300557_000011-scaled.jpg","datePublished":"2026-09-25T14:13:37+00:00","dateModified":"2026-09-25T17:22:14+00:00","author":{"@id":"https:\/\/legrandcontinent.eu\/fr\/#\/schema\/person\/e4f85f52220fc7b4cc3cfdb86d88e833"},"breadcrumb":{"@id":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#breadcrumb"},"inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/"]}]},{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#primaryimage","url":"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/SIPA_01300557_000011-scaled.jpg","contentUrl":"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/SIPA_01300557_000011-scaled.jpg","width":2560,"height":1706,"caption":"Sam Altman, Chief Executive Officer, OpenAI speaks during Security Council meeting on Artificial intelligence and international security at UN Headquarters in New York, NY on September 23, 2026. (Photo by Lev Radin\/Pacific Press) - LevRadin_092426(2) \/\/PACIFICPRESS_LevRadin_092426(2)\/Credit:Lev Radin\/PACIFIC PRESS\/SIPA\/2609241638"},{"@type":"BreadcrumbList","@id":"https:\/\/legrandcontinent.eu\/fr\/2026\/09\/25\/washington-refuse-quanthropic-et-openai-soumettent-leurs-derniers-modeles-a-une-evaluation-independante\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Accueil","item":"https:\/\/legrandcontinent.eu\/fr\/"},{"@type":"ListItem","position":2,"name":"Washington refuse qu\u2019Anthropic et OpenAI soumettent leurs derniers mod\u00e8les \u00e0 une \u00e9valuation ind\u00e9pendante"}]},{"@type":"WebSite","@id":"https:\/\/legrandcontinent.eu\/fr\/#website","url":"https:\/\/legrandcontinent.eu\/fr\/","name":"Le Grand Continent","description":"L&#039;\u00e9chelle pertinente","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/legrandcontinent.eu\/fr\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"fr-FR"},{"@type":"Person","@id":"https:\/\/legrandcontinent.eu\/fr\/#\/schema\/person\/e4f85f52220fc7b4cc3cfdb86d88e833","name":"Marin Saillofest","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/legrandcontinent.eu\/fr\/#\/schema\/person\/image\/","url":"https:\/\/secure.gravatar.com\/avatar\/87088698b4d967390c49ae6d4f1aa459?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/87088698b4d967390c49ae6d4f1aa459?s=96&d=mm&r=g","caption":"Marin Saillofest"}}]}},"lgc_categories":[{"id":3412,"name":"Puissances de l'IA","slug":"puissances-de-lia"}],"lgc_author":{"id":5175,"name":"Henri Taurel","slug":"henri-taurel","description":"<div id=\"fullDesc\" class=\"max-h-0 overflow-hidden text-styled transition-all duration-500\">\r\n<div class=\"pt-2s\">Henri Taurel est le pseudonyme d\u2019un ing\u00e9nieur sp\u00e9cialiste de l\u2019IA.<\/div>\r\n<\/div>","first_name":"Henri","last_name":"Taurel","picture":null},"feature_image_summary":{"url":"https:\/\/legrandcontinent.eu\/fr\/wp-content\/uploads\/sites\/2\/2026\/09\/SIPA_01300557_000011-scaled.jpg","width":2560,"height":1706,"alt":""},"feature_image_cover":null,"lgc_related_posts":[364227,364066,362431],"term_position_data":null,"lgc_audio":{"available":false,"generated_at":""},"_links":{"self":[{"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/posts\/359810","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/users\/2622"}],"replies":[{"embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/comments?post=359810"}],"version-history":[{"count":9,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/posts\/359810\/revisions"}],"predecessor-version":[{"id":359890,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/posts\/359810\/revisions\/359890"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/media\/359814"}],"wp:attachment":[{"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/media?parent=359810"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/categories?post=359810"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/tags?post=359810"},{"taxonomy":"staff","embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/staff?post=359810"},{"taxonomy":"editorial_format","embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/editorial_format?post=359810"},{"taxonomy":"serie","embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/serie?post=359810"},{"taxonomy":"audience","embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/audience?post=359810"},{"taxonomy":"geo","embeddable":true,"href":"https:\/\/legrandcontinent.eu\/fr\/wp-json\/wp\/v2\/geo?post=359810"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}