{"id":838,"date":"2026-04-11T23:02:48","date_gmt":"2026-04-11T23:02:48","guid":{"rendered":"https:\/\/caisi.local\/research_project\/ia-systemes-multi-agents-logistique\/"},"modified":"2026-04-20T23:41:14","modified_gmt":"2026-04-20T23:41:14","slug":"ia-systemes-multi-agents-logistique","status":"publish","type":"research_project","link":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/","title":{"rendered":"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML"},"content":{"rendered":"

L\u2019\u00e9valuation des vuln\u00e9rabilit\u00e9s des grands mod\u00e8les de langage (GML) est devenue un domaine cl\u00e9 de la recherche sur la s\u00e9curit\u00e9 de l\u2019IA. Gauthier Gidel, titulaire de chaire en IA Canada-CIFAR, propose une m\u00e9thode novatrice, plus efficace et automatis\u00e9e, pour trouver les vuln\u00e9rabilit\u00e9s dans les GML. En faisant appel \u00e0 l\u2019optimisation et en empruntant des m\u00e9thodes issues d\u2019attaques antagonistes fond\u00e9es sur les images, ce projet a pour but de fournir un mod\u00e8le efficace d\u2019attaque automatique. Cela permettra aux concepteurs et conceptrices de mod\u00e8les d\u2019am\u00e9liorer les \u00e9valuations et l\u2019entra\u00eenement des GML, de mesurer leur vuln\u00e9rabilit\u00e9, ainsi que de les rendre plus s\u00fbrs et plus robustes.<\/p>\n","protected":false},"excerpt":{"rendered":"

Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.<\/p>\n","protected":false},"featured_media":647,"template":"","ai-safety-technique":[61,60],"application-area":[76],"project-type":[85],"research-theme":[92],"class_list":["post-838","research_project","type-research_project","status-publish","has-post-thumbnail","hentry","ai-safety-technique-commandabilite-et-alignement","ai-safety-technique-robustesse-antagoniste-et-methode-de-lequipe-rouge","application-area-grands-modeles-de-langage","project-type-projet-catalyseur","research-theme-ia-fiable-et-interpretable"],"acf":[],"yoast_head":"\nLa robustesse antagoniste de la s\u00e9curit\u00e9 des GML<\/title>\n<meta name=\"description\" content=\"Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML\" \/>\n<meta property=\"og:description\" content=\"Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/\" \/>\n<meta property=\"og:site_name\" content=\"CAISI\" \/>\n<meta property=\"article:modified_time\" content=\"2026-04-20T23:41:14+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/aisafety.ca\/wp-content\/uploads\/2026\/04\/projects-catalyst-colour-3.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1920\" \/>\n\t<meta property=\"og:image:height\" content=\"1080\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:title\" content=\"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML\" \/>\n<meta name=\"twitter:description\" content=\"Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.\" \/>\n<meta name=\"twitter:image\" content=\"https:\/\/aisafety.ca\/wp-content\/uploads\/2026\/04\/projects-catalyst-colour-3.jpg\" \/>\n<meta name=\"twitter:label1\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data1\" content=\"1 minute\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/research_project\\\/ia-systemes-multi-agents-logistique\\\/\",\"url\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/research_project\\\/ia-systemes-multi-agents-logistique\\\/\",\"name\":\"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/research_project\\\/ia-systemes-multi-agents-logistique\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/research_project\\\/ia-systemes-multi-agents-logistique\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/aisafety.ca\\\/wp-content\\\/uploads\\\/2026\\\/04\\\/projects-catalyst-colour-3.jpg\",\"datePublished\":\"2026-04-11T23:02:48+00:00\",\"dateModified\":\"2026-04-20T23:41:14+00:00\",\"description\":\"Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/research_project\\\/ia-systemes-multi-agents-logistique\\\/#breadcrumb\"},\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/aisafety.ca\\\/fr\\\/research_project\\\/ia-systemes-multi-agents-logistique\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/research_project\\\/ia-systemes-multi-agents-logistique\\\/#primaryimage\",\"url\":\"https:\\\/\\\/aisafety.ca\\\/wp-content\\\/uploads\\\/2026\\\/04\\\/projects-catalyst-colour-3.jpg\",\"contentUrl\":\"https:\\\/\\\/aisafety.ca\\\/wp-content\\\/uploads\\\/2026\\\/04\\\/projects-catalyst-colour-3.jpg\",\"width\":1920,\"height\":1080,\"caption\":\"Abstract background with geometric shapes\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/research_project\\\/ia-systemes-multi-agents-logistique\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/#website\",\"url\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/\",\"name\":\"CAISI\",\"description\":\"\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/aisafety.ca\\\/fr\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"fr-FR\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML","description":"Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.","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:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/","og_locale":"fr_FR","og_type":"article","og_title":"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML","og_description":"Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.","og_url":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/","og_site_name":"CAISI","article_modified_time":"2026-04-20T23:41:14+00:00","og_image":[{"width":1920,"height":1080,"url":"https:\/\/aisafety.ca\/wp-content\/uploads\/2026\/04\/projects-catalyst-colour-3.jpg","type":"image\/jpeg"}],"twitter_card":"summary_large_image","twitter_title":"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML","twitter_description":"Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.","twitter_image":"https:\/\/aisafety.ca\/wp-content\/uploads\/2026\/04\/projects-catalyst-colour-3.jpg","twitter_misc":{"Dur\u00e9e de lecture estim\u00e9e":"1 minute"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/","url":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/","name":"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML","isPartOf":{"@id":"https:\/\/aisafety.ca\/fr\/#website"},"primaryImageOfPage":{"@id":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/#primaryimage"},"image":{"@id":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/#primaryimage"},"thumbnailUrl":"https:\/\/aisafety.ca\/wp-content\/uploads\/2026\/04\/projects-catalyst-colour-3.jpg","datePublished":"2026-04-11T23:02:48+00:00","dateModified":"2026-04-20T23:41:14+00:00","description":"Mettre au point un mod\u00e8le d\u2019attaque automatis\u00e9 performant pour optimiser l\u2019\u00e9valuation et l\u2019entra\u00eenement des grands mod\u00e8les de langage en vue de les rendre plus s\u00fbrs et plus robustes.","breadcrumb":{"@id":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/#breadcrumb"},"inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/"]}]},{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/#primaryimage","url":"https:\/\/aisafety.ca\/wp-content\/uploads\/2026\/04\/projects-catalyst-colour-3.jpg","contentUrl":"https:\/\/aisafety.ca\/wp-content\/uploads\/2026\/04\/projects-catalyst-colour-3.jpg","width":1920,"height":1080,"caption":"Abstract background with geometric shapes"},{"@type":"BreadcrumbList","@id":"https:\/\/aisafety.ca\/fr\/research_project\/ia-systemes-multi-agents-logistique\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/aisafety.ca\/fr\/"},{"@type":"ListItem","position":2,"name":"La robustesse antagoniste de la s\u00e9curit\u00e9 des GML"}]},{"@type":"WebSite","@id":"https:\/\/aisafety.ca\/fr\/#website","url":"https:\/\/aisafety.ca\/fr\/","name":"CAISI","description":"","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/aisafety.ca\/fr\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"fr-FR"}]}},"_links":{"self":[{"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/research_project\/838","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/research_project"}],"about":[{"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/types\/research_project"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/media\/647"}],"wp:attachment":[{"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/media?parent=838"}],"wp:term":[{"taxonomy":"ai-safety-technique","embeddable":true,"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/ai-safety-technique?post=838"},{"taxonomy":"application-area","embeddable":true,"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/application-area?post=838"},{"taxonomy":"project-type","embeddable":true,"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/project-type?post=838"},{"taxonomy":"research-theme","embeddable":true,"href":"https:\/\/aisafety.ca\/fr\/wp-json\/wp\/v2\/research-theme?post=838"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}