{"id":1145874,"date":"2025-08-07T15:09:08","date_gmt":"2025-08-07T22:09:08","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-project&#038;p=1145874"},"modified":"2026-04-19T23:36:21","modified_gmt":"2026-04-20T06:36:21","slug":"roma","status":"publish","type":"msr-project","link":"https:\/\/www.microsoft.com\/en-us\/research\/project\/roma\/","title":{"rendered":"Project Roma"},"content":{"rendered":"<section class=\"mb-3 moray-highlight\">\n\t<div class=\"card-img-overlay mx-lg-0\">\n\t\t<div class=\"card-background  has-background-gable-green card-background--full-bleed\">\n\t\t\t<img loading=\"lazy\" decoding=\"async\" width=\"1920\" height=\"720\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/08\/Roma-header-v3_1920x720.jpg\" class=\"attachment-full size-full\" alt=\"aerial view of Rome\" style=\"object-position: 49% 24%\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/08\/Roma-header-v3_1920x720.jpg 1920w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/08\/Roma-header-v3_1920x720-300x113.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/08\/Roma-header-v3_1920x720-1024x384.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/08\/Roma-header-v3_1920x720-768x288.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/08\/Roma-header-v3_1920x720-1536x576.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/08\/Roma-header-v3_1920x720-1600x600.jpg 1600w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/08\/Roma-header-v3_1920x720-240x90.jpg 240w\" sizes=\"auto, (max-width: 1920px) 100vw, 1920px\" \/>\t\t<\/div>\n\t\t<!-- Foreground -->\n\t\t<div class=\"card-foreground d-flex mt-md-n5 my-lg-5 px-g px-lg-0\">\n\t\t\t<!-- Container -->\n\t\t\t<div class=\"container d-flex mt-md-n5 my-lg-5 \">\n\t\t\t\t<!-- Card wrapper -->\n\t\t\t\t<div class=\"w-100 w-lg-col-5\">\n\t\t\t\t\t<!-- Card -->\n\t\t\t\t\t<div class=\"card material-md-card py-5 px-md-5\">\n\t\t\t\t\t\t<div class=\"card-body \">\n\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\n\n<h1 class=\"wp-block-heading\" id=\"project-roma\">Project Roma<\/h1>\n\n\n\n<p>Deterministic security for AI agents<\/p>\n\n\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\t<\/div>\n<\/section>\n\n\n\n\n\n<h2 class=\"wp-block-heading\" id=\"project-roma-deterministic-security-guarantees-for-ai-agents-1\">Project Roma: Deterministic security for AI agents<\/h2>\n\n\n\n<p>AI agents perform consequential actions while processing data from various sources, including trusted collaborators and the public Web. It is crucial that AI agents handle this data with care: confidential data must be adequately protected, and untrusted data must not derail the agent&#8217;s behavior. However, AI agents rely on models that can behave unpredictably and are susceptible to manipulation. This makes them vulnerable to attacks such as indirect prompt injection attacks, which can steal and corrupt data.&nbsp;<\/p>\n\n\n\n<p>Project Roma aims to create a system-level, deterministic protective layer that hardens AI agents, providing strong security and confidentiality guarantees even when the models they use misbehave.<\/p>\n\n\n\n<div style=\"height:30px\" aria-hidden=\"true\" class=\"wp-block-spacer\"><\/div>\n\n\n","protected":false},"excerpt":{"rendered":"<p>Deterministic security for AI agents AI agents perform consequential actions while processing data from various sources, including trusted collaborators and the public Web. It is crucial that AI agents handle this data with care: confidential data must be adequately protected, and untrusted data must not derail the agent&#8217;s behavior. However, AI agents rely on models [&hellip;]<\/p>\n","protected":false},"featured_media":1147556,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","footnotes":""},"research-area":[13556,13558],"msr-locale":[268875],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1145874","msr-project","type-msr-project","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-research-area-security-privacy-cryptography","msr-locale-en_us","msr-archive-status-active"],"msr_project_start":"","related-publications":[1146435,1146437,1162338],"related-downloads":[],"related-videos":[],"related-groups":[559983],"related-events":[],"related-opportunities":[],"related-posts":[],"related-articles":[],"tab-content":[],"slides":[],"related-researchers":[{"type":"user_nicename","display_name":"Manuel Costa","user_id":32794,"people_section":"Section name 0","alias":"manuelc"},{"type":"user_nicename","display_name":"Aashish Kolluri","user_id":43955,"people_section":"Section name 0","alias":"t-akolluri"},{"type":"user_nicename","display_name":"Boris K&ouml;pf","user_id":37857,"people_section":"Section name 0","alias":"bokoepf"},{"type":"user_nicename","display_name":"Andrew Paverd","user_id":37902,"people_section":"Section name 0","alias":"anpaverd"},{"type":"guest","display_name":"Mark Russinovich","user_id":591880,"people_section":"Section name 0","alias":""},{"type":"user_nicename","display_name":"Ahmed Salem","user_id":43959,"people_section":"Section name 0","alias":"ahmsalem"},{"type":"user_nicename","display_name":"Rishi Sharma","user_id":44144,"people_section":"Section name 0","alias":"rishisharma"},{"type":"user_nicename","display_name":"Shruti Tople","user_id":39003,"people_section":"Section name 0","alias":"shtople"},{"type":"user_nicename","display_name":"Lukas Wutschitz","user_id":38775,"people_section":"Section name 0","alias":"luwutsch"},{"type":"user_nicename","display_name":"Santiago Zanella-B\u00e9guelin","user_id":33518,"people_section":"Section name 0","alias":"santiago"}],"msr_research_lab":[],"msr_impact_theme":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1145874","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-project"}],"version-history":[{"count":10,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1145874\/revisions"}],"predecessor-version":[{"id":1168918,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1145874\/revisions\/1168918"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1147556"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1145874"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1145874"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1145874"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1145874"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1145874"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}