{"id":222862,"date":"2026-11-04T05:39:54","date_gmt":"2026-11-04T05:39:54","guid":{"rendered":"https:\/\/10pearls.com\/uk\/blog\/\/"},"modified":"2026-11-04T07:59:20","modified_gmt":"2026-11-04T07:59:20","slug":"securing-ai-agents","status":"publish","type":"post","link":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/","title":{"rendered":"Securing AI Agents: 4 New Risks and 7 Controls"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"222862\" class=\"elementor elementor-222862\" data-elementor-post-type=\"post\">\n\t\t\t\t<div class=\"elementor-element elementor-element-4e550181 e-flex e-con-boxed e-con e-parent\" data-id=\"4e550181\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-6a742e96 elementor-widget__width-initial indigo-h1 elementor-widget elementor-widget-heading\" data-id=\"6a742e96\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h1 class=\"elementor-heading-title elementor-size-default\">Securing AI Agents: Autonomous Intelligence Demands a New Security Mindset\u00a0<\/h1>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-18e5d2b7 elementor-icon-list--layout-inline elementor-list-item-link-full_width elementor-widget elementor-widget-icon-list\" data-id=\"18e5d2b7\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-list.default\">\n\t\t\t\t\t\t\t<ul class=\"elementor-icon-list-items elementor-inline-items\">\n\t\t\t\t\t\t\t<li class=\"elementor-icon-list-item elementor-inline-item\">\n\t\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-icon\">\n\t\t\t\t\t\t\t<svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"81\" height=\"81\" viewBox=\"0 0 81 81\" fill=\"none\"><path fill-rule=\"evenodd\" clip-rule=\"evenodd\" d=\"M41.9662 0.167969C54.7101 0.167969 65.9976 6.21222 73.207 15.5335C70.8038 14.2955 68.1822 13.4216 65.5606 12.8391L64.2498 12.6206L62.5749 12.4749H57.9871L56.1665 12.6206C45.1704 14.514 36.8686 22.2331 33.8829 32.6467C33.2275 34.9771 32.9362 37.3074 32.9362 39.7105C32.9362 44.5168 34.1014 49.2502 36.5773 53.4011C40.7282 60.392 48.0104 65.4168 56.5306 66.5819L58.3512 66.7276H62.7934L64.6139 66.5819C67.5996 66.0722 70.4397 65.1255 73.0613 63.7419C65.5606 73.4272 53.5449 79.5443 40.2184 78.9617C19.8282 78.0879 3.29751 61.5572 2.49646 41.1669C1.69541 18.7377 19.6825 0.167969 41.9662 0.167969ZM21.8672 65.6353H30.8972V14.1499H12.4003L17.862 23.4711H21.8672V65.6353Z\" fill=\"black\"><\/path><path fill-rule=\"evenodd\" clip-rule=\"evenodd\" d=\"M60.5396 20.9219C70.8803 20.9219 79.2549 29.2964 79.2549 39.6372C79.2549 49.978 70.8803 58.3526 60.5396 58.3526C50.1988 58.3526 41.8242 49.978 41.8242 39.6372C41.8242 29.2964 50.1988 20.9219 60.5396 20.9219Z\" fill=\"black\"><\/path><\/svg>\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-text\">10Pearls Editorial Team<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t<li class=\"elementor-icon-list-item elementor-inline-item\">\n\t\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-icon\">\n\t\t\t\t\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-circle\" viewBox=\"0 0 512 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M256 8C119 8 8 119 8 256s111 248 248 248 248-111 248-248S393 8 256 8z\"><\/path><\/svg>\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-text\">10 min read<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n\t\t\t\t\t\t<\/ul>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-12c78bd4 e-flex e-con-boxed e-con e-parent\" data-id=\"12c78bd4\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t<div class=\"elementor-element elementor-element-41285805 e-con-full e-flex e-con e-child\" data-id=\"41285805\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-506f3451 elementor-widget elementor-widget-heading\" data-id=\"506f3451\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">Summary<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-28521ea8 elementor-widget elementor-widget-text-editor\" data-id=\"28521ea8\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\tIn this blog, we go over how AI agents disrupt the conventional security model and introduce knew threat dimensions, and how \nsecurity fundamentals should be reinterpreted for agents. We also discuss the security controls that should be engineered into \nagent architectures.\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-9b53d62 e-con-full e-flex e-con e-child\" data-id=\"9b53d62\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t<div class=\"elementor-element elementor-element-407a53ba e-con-full e-flex e-con e-child\" data-id=\"407a53ba\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t<div class=\"elementor-element elementor-element-3f6060f6 section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"3f6060f6\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>Artificial intelligence has moved beyond models and copilots. Teams are now deploying AI agents. It\u2019s not merely capability expansion but rather the evolution of how AI can transform operational models and processes of different businesses. But this transformation, operational efficiencies, and other benefits that these agents promise, are not without a cost.<\/p><p>AI agents are systems capable of planning, executing, adapting, and acting both inside and outside enterprise environments autonomously. The degree of autonomy these agents have varies based on agent-specific permission, the workflow in which the agent operates, the underlying agentic AI framework, and agent orchestration. Even with humans in the loop, agentic autonomy represents a significant shift that introduces new risks we are already observing across deployments:<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-22858d15 e-grid e-con-full section-head-margin-bottom e-con e-child\" data-id=\"22858d15\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t<div class=\"elementor-element elementor-element-1d4d02ce card-shadow elementor-widget elementor-widget-icon-box\" data-id=\"1d4d02ce\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tA <strong>development agent<\/strong> <a href=\"https:\/\/fortune.com\/2025\/07\/23\/ai-coding-tool-replit-wiped-database-called-it-a-catastrophic-failure\/\" target=\"_blank\" rel=\"nofollow\">deleted a live production database<\/a> during an automated workflow.\u00a0\u00a0\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6ef20f98 card-shadow elementor-widget elementor-widget-icon-box\" data-id=\"6ef20f98\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tAn <strong>AI development tool<\/strong> <a href=\"https:\/\/www.techradar.com\/ai-platforms-assistants\/googles-antigravity-ai-deleted-a-developers-drive-and-then-apologized\" target=\"_blank\" rel=\"nofollow\">wiped a developer\u2019s drive<\/a> after misinterpreting a system instruction\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7e8de4ae card-shadow elementor-widget elementor-widget-icon-box\" data-id=\"7e8de4ae\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tA <strong>productivity assistant<\/strong> <a href=\"https:\/\/techcrunch.com\/2026\/02\/18\/microsoft-says-office-bug-exposed-customers-confidential-emails-to-copilot-ai\/\" target=\"_blank\" rel=\"nofollow\">exposed confidential emails<\/a> due to a permissions flaw.\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6cfcb591 card-shadow elementor-widget elementor-widget-icon-box\" data-id=\"6cfcb591\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tA <strong>containerized AI assistant<\/strong> vulnerability enabled <a href=\"https:\/\/www.securityweek.com\/dockerdash-flaw-in-docker-ai-assistant-leads-to-rce-data-theft\/\" target=\"_blank\" rel=\"nofollow\">malicious instruction injection and potential remote code execution<\/a>.\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1f1b1988 section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"1f1b1988\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>These incidents are not simply \u201cAI bugs\u201d or model-level limitations. They are signals that AI agents operate under a fundamentally different and more dynamic risk model than traditional software or even flawed human developers.\u00a0\u00a0<\/p><p>The lesson is clear: AI agents must be engineered, integrated, orchestrated, and governed with security as a first-class design constraint, not an afterthought.\u00a0\u00a0<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-eb72a2d indigo-h2 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"eb72a2d\" data-element_type=\"widget\" data-e-type=\"widget\" id=\"01\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">How AI agents differ from traditional applications<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-65a24431 section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"65a24431\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>Traditional applications execute deterministic instructions. If there is a flaw, it exists in known code paths and can be patched. Vulnerabilities can be identified or predicted with optimal accuracy through code assessment and past performance evaluation.<\/p><p>AI agents behave differently from traditional systems, regardless of their level of autonomy. They are essentially goal-driven systems that:<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1e77e4e3 elementor-icon-list--layout-inline section-head-margin-bottom elementor-list-item-link-full_width elementor-widget elementor-widget-icon-list\" data-id=\"1e77e4e3\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-list.default\">\n\t\t\t\t\t\t\t<ul class=\"elementor-icon-list-items elementor-inline-items\">\n\t\t\t\t\t\t\t<li class=\"elementor-icon-list-item elementor-inline-item\">\n\t\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-icon\">\n\t\t\t\t\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-circle\" viewBox=\"0 0 512 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M256 8C119 8 8 119 8 256s111 248 248 248 248-111 248-248S393 8 256 8z\"><\/path><\/svg>\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-text\">Interpret high-level objectives\u00a0\u00a0<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t<li class=\"elementor-icon-list-item elementor-inline-item\">\n\t\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-icon\">\n\t\t\t\t\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-circle\" viewBox=\"0 0 512 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M256 8C119 8 8 119 8 256s111 248 248 248 248-111 248-248S393 8 256 8z\"><\/path><\/svg>\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-text\">Generate multi-step plans\u00a0\u00a0<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t<li class=\"elementor-icon-list-item elementor-inline-item\">\n\t\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-icon\">\n\t\t\t\t\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-circle\" viewBox=\"0 0 512 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M256 8C119 8 8 119 8 256s111 248 248 248 248-111 248-248S393 8 256 8z\"><\/path><\/svg>\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-text\">Adapt when they encounter obstacles<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t<li class=\"elementor-icon-list-item elementor-inline-item\">\n\t\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-icon\">\n\t\t\t\t\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-circle\" viewBox=\"0 0 512 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M256 8C119 8 8 119 8 256s111 248 248 248 248-111 248-248S393 8 256 8z\"><\/path><\/svg>\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-text\">Reflect or amplify model-level variability if misconfigured\u00a0<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t<li class=\"elementor-icon-list-item elementor-inline-item\">\n\t\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-icon\">\n\t\t\t\t\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-circle\" viewBox=\"0 0 512 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M256 8C119 8 8 119 8 256s111 248 248 248 248-111 248-248S393 8 256 8z\"><\/path><\/svg>\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-text\">Invoke tools across system boundaries\u00a0\u00a0<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t<li class=\"elementor-icon-list-item elementor-inline-item\">\n\t\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-icon\">\n\t\t\t\t\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-circle\" viewBox=\"0 0 512 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M256 8C119 8 8 119 8 256s111 248 248 248 248-111 248-248S393 8 256 8z\"><\/path><\/svg>\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t<span class=\"elementor-icon-list-text\">Dynamically chain actions across steps\u00a0<\/span>\n\t\t\t\t\t\t\t\t\t<\/li>\n\t\t\t\t\t\t<\/ul>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6e0f3803 section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"6e0f3803\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>This means security teams are no longer able to govern what the code does \u2013 they must govern how AI agents pursue objectives, individually, and as orchestrated systems.<\/p><p>The governance must be flexible enough to reduce risk without undermining agentic autonomy and capabilities. It&#8217;s no longer enough to treat the entire security posture as a living, evolving system with overarching governance and policy constraints alone. The controls must account for agent-level autonomy, orchestration frameworks, and multi-agent configurations.<\/p><p>The difference means our traditional threat models will no longer get the job done. Even AI-aware threat models don\u2019t fully account for agentic autonomy.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-35b61174 section-head-margin-bottom elementor-widget elementor-widget-image\" data-id=\"35b61174\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img loading=\"lazy\" decoding=\"async\" width=\"860\" height=\"455\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/difference-between-ai-agents-and-traditional-applications-1.webp\" class=\"attachment-large size-large wp-image-222870\" alt=\"\" srcset=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/difference-between-ai-agents-and-traditional-applications-1.webp 860w, https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/difference-between-ai-agents-and-traditional-applications-1-300x159.webp 300w, https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/difference-between-ai-agents-and-traditional-applications-1-768x406.webp 768w\" sizes=\"auto, (max-width: 860px) 100vw, 860px\" loading=\"lazy\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-54ebb7c6 indigo-h2 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"54ebb7c6\" data-element_type=\"widget\" data-e-type=\"widget\" id=\"02\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">How AI agents disrupt traditional \nsecurity models<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-38ad8ef1 section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"38ad8ef1\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>Enterprise security has traditionally relied on a predictable chain of behavior from an attacker: intrusion, privilege escalation, lateral movement, and exfiltration. Meanwhile, insider risk has been managed through access controls, identity governance, and monitoring of anomalous behavior within trusted boundaries.<\/p><p>AI agents are introducing new dimensions to that model. They inherit insider-level access and autonomy but operate with machine-scale speed and opaque reasoning, amplifying the impact of traditional insider risk.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-2cff6d61 e-con-full section-head-margin-bottom e-flex e-con e-child\" data-id=\"2cff6d61\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t<div class=\"elementor-element elementor-element-21834391 e-con-full e-flex e-con e-child\" data-id=\"21834391\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t<div class=\"elementor-element elementor-element-54acf6dc counter-icon elementor-view-default elementor-widget elementor-widget-icon\" data-id=\"54acf6dc\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-wrapper\">\n\t\t\t<div class=\"elementor-icon\">\n\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-star\" viewBox=\"0 0 576 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M259.3 17.8L194 150.2 47.9 171.5c-26.2 3.8-36.7 36.1-17.7 54.6l105.7 103-25 145.5c-4.5 26.3 23.2 46 46.4 33.7L288 439.6l130.7 68.7c23.2 12.2 50.9-7.4 46.4-33.7l-25-145.5 105.7-103c19-18.5 8.5-50.8-17.7-54.6L382 150.2 316.7 17.8c-11.7-23.6-45.6-23.9-57.4 0z\"><\/path><\/svg>\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-a164bee e-flex e-con-boxed e-con e-child\" data-id=\"a164bee\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-760fe2b9 indigo-h3 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"760fe2b9\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">The black box problem<\/h3>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-377dfb46 elementor-widget elementor-widget-text-editor\" data-id=\"377dfb46\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>Agents operate with internal reasoning processes that are not fully observable (and sometimes not observable at all). It\u2019s not an agent design or deployment flaw that can be rectified with the right architecture or configuration, but instead a natural consequence of how underlying LLMs are built and trained.<\/p><p>With very limited exceptions, an LLM\u2019s \u201cthought process\u201d is a black box. We often see only the final action of the agents powered by these LLMs &#8211; not the decision path that produced it. This limits early detection and complicates post-incident analysis. Also, most agents rely on foundation models, all of which suffer from the same limitations in terms of observability.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-29844056 e-flex e-con-boxed e-con e-child\" data-id=\"29844056\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-310b003f counter-icon elementor-view-default elementor-widget elementor-widget-icon\" data-id=\"310b003f\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-wrapper\">\n\t\t\t<div class=\"elementor-icon\">\n\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-star\" viewBox=\"0 0 576 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M259.3 17.8L194 150.2 47.9 171.5c-26.2 3.8-36.7 36.1-17.7 54.6l105.7 103-25 145.5c-4.5 26.3 23.2 46 46.4 33.7L288 439.6l130.7 68.7c23.2 12.2 50.9-7.4 46.4-33.7l-25-145.5 105.7-103c19-18.5 8.5-50.8-17.7-54.6L382 150.2 316.7 17.8c-11.7-23.6-45.6-23.9-57.4 0z\"><\/path><\/svg>\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-7ca4cbf0 e-flex e-con-boxed e-con e-child\" data-id=\"7ca4cbf0\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-4972ec6d indigo-h3 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"4972ec6d\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Adaptive behavior &amp; goal drift <\/h3>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-72050e23 elementor-widget elementor-widget-text-editor\" data-id=\"72050e23\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>An agent tasked with \u201creducing support backlog\u201d may begin closing tickets aggressively or bypass validation checks if they interpret efficiency as success. Whatever it interprets as the goal, it optimizes, and the AI agent\u2019s optimization may not match the business intent.<\/p><p>In security-sensitive workflows, this can lead to control bypass, misclassification of risk, or unresolved vulnerabilities being treated as remediated.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-32443835 e-flex e-con-boxed e-con e-child\" data-id=\"32443835\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-2809ee72 counter-icon elementor-view-default elementor-widget elementor-widget-icon\" data-id=\"2809ee72\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-wrapper\">\n\t\t\t<div class=\"elementor-icon\">\n\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-star\" viewBox=\"0 0 576 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M259.3 17.8L194 150.2 47.9 171.5c-26.2 3.8-36.7 36.1-17.7 54.6l105.7 103-25 145.5c-4.5 26.3 23.2 46 46.4 33.7L288 439.6l130.7 68.7c23.2 12.2 50.9-7.4 46.4-33.7l-25-145.5 105.7-103c19-18.5 8.5-50.8-17.7-54.6L382 150.2 316.7 17.8c-11.7-23.6-45.6-23.9-57.4 0z\"><\/path><\/svg>\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-c9d23bb e-flex e-con-boxed e-con e-child\" data-id=\"c9d23bb\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-415c53af indigo-h3 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"415c53af\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Autonomous tool use\u00a0<\/h3>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-4cf6a743 elementor-widget elementor-widget-text-editor\" data-id=\"4cf6a743\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>Part of the power of these agents is that they can work without humans in the loop, to make processes more efficient, save time, and reduce the number of human errors. They can invoke APIs, execute commands, or interact with other enterprise and external systems.<\/p><p>Multiple agents work in tandem for complex workflows, with their collective tooling and access permissions compounding the risk and attack surface. This expands both insider and externally influenced risk.<\/p><p>Attackers can poison the tools to harm or manipulate agents, and agents can expose sensitive data by implicitly trusting all tools.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-748bc6c e-con-full e-flex e-con e-child\" data-id=\"748bc6c\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t<div class=\"elementor-element elementor-element-1922eafa counter-icon elementor-view-default elementor-widget elementor-widget-icon\" data-id=\"1922eafa\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-wrapper\">\n\t\t\t<div class=\"elementor-icon\">\n\t\t\t<svg aria-hidden=\"true\" class=\"e-font-icon-svg e-fas-star\" viewBox=\"0 0 576 512\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\"><path d=\"M259.3 17.8L194 150.2 47.9 171.5c-26.2 3.8-36.7 36.1-17.7 54.6l105.7 103-25 145.5c-4.5 26.3 23.2 46 46.4 33.7L288 439.6l130.7 68.7c23.2 12.2 50.9-7.4 46.4-33.7l-25-145.5 105.7-103c19-18.5 8.5-50.8-17.7-54.6L382 150.2 316.7 17.8c-11.7-23.6-45.6-23.9-57.4 0z\"><\/path><\/svg>\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-83f631d e-flex e-con-boxed e-con e-child\" data-id=\"83f631d\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-6432eb3e indigo-h3 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"6432eb3e\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Continuous learning &amp; context evolution<\/h3>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-98d69dc elementor-widget elementor-widget-text-editor\" data-id=\"98d69dc\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>Unlike static applications, agent behavior evolves. The evolution may be due to changes in prompts, data, or environmental context. They may also reflect underlying model variability in their behavior. This means that the assumptions security teams make today may not hold true in the future.<\/p><p>Stifling this evolution for security undermines agentic benefits, so controls must adapt accordingly. However, poorly designed controls can introduce unnecessary complexity and operational overhead.<\/p><p>Traditional security teams look for chains of behavior, assuming adversarial intent and insider risks. In contrast, agentic risk emerges from a combination of misaligned optimizations, over- or under-permissive access controls, continuous adaptability, and model-level limitations.<\/p><p>AI agents don\u2019t just introduce new risks\u2014they change where risk originates and how it spreads across systems.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-764b3663 indigo-h2 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"764b3663\" data-element_type=\"widget\" data-e-type=\"widget\" id=\"03\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">Reinterpreting security fundamentals for \nAI agents <\/h2>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-19921dad section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"19921dad\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>Security teams should consider AI agents like employees with system-level access. But unlike human insiders, AI agents:<\/p><ul><li><span style=\"white-space: pre-wrap; background-color: #ffffff;\">Operate at machine speed\u00a0\u00a0<\/span><\/li><li><span style=\"white-space: pre-wrap; background-color: #ffffff;\">Lack holistic contextual judgment beyond their provided inputs\u00a0<\/span><\/li><li><span style=\"white-space: pre-wrap; background-color: #ffffff;\">Have no innate understanding of organizational or situational risk tolerance\u00a0\u00a0<\/span><\/li><li><span style=\"white-space: pre-wrap; background-color: #ffffff;\">Don\u2019t inherently adhere to policy or ethical constraints unless explicitly configured<\/span><\/li><\/ul><p>If a mid-level engineer would not be granted the privileges to a full production database, or to delete filesystems without oversight, neither should an autonomous agent.<\/p><p>The difference is that if overly broad permissions are in place, AI agents can leverage those privileges instantly, repeatedly, and without fatigue, concern, or a true understanding of the business goals. With underspecified permissions and objectives, they tend to prioritize narrow goals over broader business objectives, triggering unintended outcomes.<\/p><p>The core risks security professionals focus on remain the same, but must be re-interpreted for AI agents:<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-345a6c3f e-con-full e-grid section-head-margin-bottom e-con e-child\" data-id=\"345a6c3f\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-499debc7 elementor-widget elementor-widget-icon-box\" data-id=\"499debc7\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tConfidentiality\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tAgents can surface or transmit sensitive data beyond intended boundaries. In a multi-agent configuration, this includes unintended inter-agent data sharing across boundaries.  \t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-b284f27 elementor-widget elementor-widget-icon-box\" data-id=\"b284f27\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tIntegrity\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tAgents can modify, corrupt, or destroy existing data or generate new data in pursuit of an interpreted objective. They can also introduce inaccurate or unverified data derived from underlying model outputs. \t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6902eaf1 elementor-widget elementor-widget-icon-box\" data-id=\"6902eaf1\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tAvailability\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tAgents can inadvertently delete or disable systems through misaligned execution. They may also modify or disrupt system workflows in pursuit of efficiency, impacting system availability. \t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-44548c38 section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"44548c38\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\tThese are not hypothetical concerns. They are real-world manifestations of an emerging threat surface. We are already witnessing these and a range of new cybersecurity threats stemming from misaligned agentic deployments.  \t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-723b605a indigo-h2 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"723b605a\" data-element_type=\"widget\" data-e-type=\"widget\" id=\"04\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">Engineering AI agents securely  <\/h2>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7973330c section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"7973330c\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\tSecure AI agent development requires extending DevSecOps principles into agent design and governance. Key pillars include:\u00a0\u00a0\t\t\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-4a97720d e-con-full e-flex e-con e-child\" data-id=\"4a97720d\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-619c059c indigo-h3 elementor-widget elementor-widget-icon-box\" data-id=\"619c059c\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tConstrained autonomy\u00a0\u00a0\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tDefine clearly whether the agent recommends actions or executes them. High-risk actions should require human verification. The autonomy can also be policy-gated and tiered based on criticality and risk. Agent actions can be tested in secure simulations before being executed in operational environments.\u00a0\u00a0\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-15b7d579 e-con-full e-flex e-con e-child\" data-id=\"15b7d579\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-327b79e9 indigo-h3 elementor-widget elementor-widget-icon-box\" data-id=\"327b79e9\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tLimited permissions\u00a0\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tBy default, agents should have fewer privileges and access permissions compared to human engineers. Least privilege is ideal. Short-lived tokens can secure agent-to-agent permissions and on-behalf-of delegation for user actions. Separating access and execution permissions and enabling just-in-time privilege escalation can support efficient agentic operations while reducing risk.\u00a0\u00a0\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-5cb1126c e-con-full e-flex e-con e-child\" data-id=\"5cb1126c\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-f3241fa indigo-h3 elementor-widget elementor-widget-icon-box\" data-id=\"f3241fa\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tEnvironment segmentation\u00a0\u00a0\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tIsolate agent execution environments while preserving workflow efficiency. Prevent cross-environment blast radius by implementing network-level isolation, restricting tool invocation, and cross-environment executions (unless policy-gated).\u00a0\u00a0\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-23da342f e-con-full e-flex e-con e-child\" data-id=\"23da342f\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-7fe0a30c indigo-h3 elementor-widget elementor-widget-icon-box\" data-id=\"7fe0a30c\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tBehavioral telemetry\u00a0\u00a0\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tLog not only actions, but reasoning traces (where available) and tool invocations. Observability must expand beyond traditional logs. Goal and outcome divergence should also be tracked to ensure alignment and active course correction. Behavioral markers of individual agents and multi-agent systems should be tracked against defined KPIs and expected outcomes.\u00a0\u00a0\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-15b7ef5a e-con-full e-flex e-con e-child\" data-id=\"15b7ef5a\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-79041e5d indigo-h3 elementor-widget elementor-widget-icon-box\" data-id=\"79041e5d\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tGovernance review\u00a0\u00a0\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tIntroduce structured risk reviews for new agent deployments. Classify agents by access, autonomy, and oversight levels before deployment. Monitoring should extend to the agent level, even in orchestrated, multi-agent deployments, to track compliance and risk at the source. A differentiated permission structure for high-impact agents and an agent registry can prevent sprawl and shadow agent use.\u00a0\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-c00d51f e-con-full e-flex e-con e-child\" data-id=\"c00d51f\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-7e405c75 indigo-h3 elementor-widget elementor-widget-icon-box\" data-id=\"7e405c75\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tTrust boundaries\u00a0\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tAll inputs and outputs from external tools must be treated as untrusted by default. Sanitize inputs, enforce tool permissions, and validate tool outputs using independent models, deterministic rules, or manual checks, as needed. Establishing tool use boundaries for multi-agent systems can minimize blast radius.\u00a0\u00a0\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-26722773 e-con-full e-flex e-con e-child\" data-id=\"26722773\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-5dc8ec8e indigo-h3 elementor-widget elementor-widget-icon-box\" data-id=\"5dc8ec8e\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon-box.default\">\n\t\t\t\t\t\t\t<div class=\"elementor-icon-box-wrapper\">\n\n\t\t\t\n\t\t\t\t\t\t<div class=\"elementor-icon-box-content\">\n\n\t\t\t\t\t\t\t\t\t<h3 class=\"elementor-icon-box-title\">\n\t\t\t\t\t\t<span  >\n\t\t\t\t\t\t\tContinuous evaluation &amp; drift monitoring\u00a0\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t<p class=\"elementor-icon-box-description\">\n\t\t\t\t\t\tConduct agent resilience and behavior evaluations for both adversarial and insider scenarios as a baseline before deployment. Then, continuously evaluate consistency during agent operation. Roll-back and kill switch mechanisms should be in place for significant data drifts, prompt drifts, and agent behavior changes.<br><br>Security must evolve from model safety evaluation to behavioral risk containment.\t\t\t\t\t<\/p>\n\t\t\t\t\n\t\t\t<\/div>\n\t\t\t\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-8533b94 section-head-margin-bottom elementor-widget elementor-widget-image\" data-id=\"8533b94\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img loading=\"lazy\" decoding=\"async\" width=\"860\" height=\"455\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/how-to-engineer-ai-agents-securely-1.webp\" class=\"attachment-large size-large wp-image-222871\" alt=\"\" srcset=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/how-to-engineer-ai-agents-securely-1.webp 860w, https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/how-to-engineer-ai-agents-securely-1-300x159.webp 300w, https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/how-to-engineer-ai-agents-securely-1-768x406.webp 768w\" sizes=\"auto, (max-width: 860px) 100vw, 860px\" loading=\"lazy\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-21cd177c indigo-h2 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"21cd177c\" data-element_type=\"widget\" data-e-type=\"widget\" id=\"05\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">The question leaders must ask\u00a0\u00a0<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7d34e1be section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"7d34e1be\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\tThe next generation of AI risk is not simply \u201cIs the model biased?\u201d or \u201cIs the output correct?\u201d It is: \t\t\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-5cc6d6d5 e-con-full e-flex e-con e-child\" data-id=\"5cc6d6d5\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-35728a2c elementor-widget elementor-widget-heading\" data-id=\"35728a2c\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<p class=\"elementor-heading-title elementor-size-default\">How much damage could this agent do based on its permissions and autonomy?<\/p>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-72897109 section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"72897109\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>That question demands a new level of rigor in AI architecture, governance, and security design. It introduces new risk dimensions and architectural constraints, especially in multi-agent deployments that are becoming the norm for complex enterprise workflows.<\/p><p>If your organization is building or deploying AI agents, now is the time to reassess your security model for agentic readiness.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7ebdd42e indigo-h2 section-title-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"7ebdd42e\" data-element_type=\"widget\" data-e-type=\"widget\" id=\"06\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">A different threat requires a different partner<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-17802de5 section-head-margin-bottom elementor-widget elementor-widget-text-editor\" data-id=\"17802de5\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t\t\t\t\t\t<p>The organizations that succeed in the AI era will not be the ones deploying agents faster than the competition. They will be those who deploy them responsibly &#8211; balancing speed with risk containment.<\/p><p>At 10Pearls, we approach AI agent development with a security-first mindset and architecture. We:<\/p><ul><li>Embed risk modeling in solution design, agent architecture, and orchestration.<\/li><li>Build governance frameworks aligned with enterprise compliance, observability, and transparency requirements.<\/li><li>Leverage secure-by-design engineering practices from development to deployment.<\/li><li>Develop continuous monitoring and containment strategies.<\/li><li>Integrate AI expertise with cybersecurity leadership.<\/li><\/ul><p>We help enterprises move from experimentation to production with confidence &#8211; designing systems that account for this new threat paradigm from day one.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-17135a51 e-con-full e-flex e-con e-child\" data-id=\"17135a51\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t<div class=\"elementor-element elementor-element-779eaf7f elementor-widget elementor-widget-heading\" data-id=\"779eaf7f\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<p class=\"elementor-heading-title elementor-size-default\">Reach out to <strong>10Pearls<\/strong> to explore how we help enterprises engineer autonomous intelligence - securely, responsibly, and at scale. <\/p>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-f5df21b e-con-full e-flex e-con e-child\" data-id=\"f5df21b\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t<div class=\"elementor-element elementor-element-15a10ce9 elementor-widget elementor-widget-heading\" data-id=\"15a10ce9\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">TABLE OF CONTENTS<\/h3>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-f5f400b elementor-widget elementor-widget-heading\" data-id=\"f5f400b\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<p class=\"elementor-heading-title elementor-size-default\"><a href=\"#01\">How AI agents differ from traditional applications\u00a0<\/a><\/p>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2a4fac56 elementor-widget elementor-widget-heading\" data-id=\"2a4fac56\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<p class=\"elementor-heading-title elementor-size-default\"><a href=\"#02\">How AI agents disrupt traditional \nsecurity models<\/a><\/p>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2b94d008 elementor-widget elementor-widget-heading\" data-id=\"2b94d008\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<p class=\"elementor-heading-title elementor-size-default\"><a href=\"#03\">Reinterpreting security fundamentals for AI agents <\/a><\/p>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-16bd4eeb elementor-widget elementor-widget-heading\" data-id=\"16bd4eeb\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<p class=\"elementor-heading-title elementor-size-default\"><a href=\"#04\">Engineering AI agents securely<\/a><\/p>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1e9869c8 elementor-widget elementor-widget-heading\" data-id=\"1e9869c8\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<p class=\"elementor-heading-title elementor-size-default\"><a href=\"#05\">The question leaders must ask\u00a0\u00a0<\/a><\/p>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-50d803f3 elementor-widget elementor-widget-heading\" data-id=\"50d803f3\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<p class=\"elementor-heading-title elementor-size-default\"><a href=\"#06\">A different threat requires a different partner  <\/a><\/p>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-2c8dd3f3 section-padding e-flex e-con-boxed e-con e-parent\" data-id=\"2c8dd3f3\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-3e49fb71 indigo-h2 section-head-margin-bottom elementor-widget elementor-widget-heading\" data-id=\"3e49fb71\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">Related blogs<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-419a7b82 elementor-widget elementor-widget-insights_section_widget\" data-id=\"419a7b82\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"insights_section_widget.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<style>\n\t\t\/* \u2500\u2500 Widget wrapper \u2500\u2500 *\/\n\t\t#isw-419a7b82 {\n\t\t\tposition: relative;\n\t\t\tpadding: 0 40px; \/* space for arrows on left\/right *\/\n\t\t\tbox-sizing: border-box;\n\t\t}\n\t\t#isw-419a7b82 .isw-outer {\n\t\t\tposition: relative;\n\t\t\toverflow: hidden;\n\t\t\tpadding: 20px 0;\n\t\t\tmargin: -20px 0;\n\t\t}\n\t\t#isw-419a7b82 .isw-track {\n\t\t\tdisplay: flex;\n\t\t\tflex-direction: row;\n\t\t\tflex-wrap: nowrap;\n\t\t\tgap: 20px;\n\t\t\ttransition: transform .5s cubic-bezier(.25,.46,.45,.94);\n\t\t\twill-change: transform;\n\t\t\talign-items: stretch;\n\t\t\tvisibility: hidden;\n\t\t}\n\t\t#isw-419a7b82 .isw-slide { flex-shrink: 0; box-sizing: border-box; display: flex; }\n\n\t\t\/* \u2500\u2500 Card \u2014 no shadow by default, only on hover \u2500\u2500 *\/\n\t\t#isw-419a7b82 .isw-card {\n\t\t\tborder-radius: 16px; overflow: hidden; width: 100%; flex: 1;\n\t\t\tposition: relative; height: 380px; background: #fff;\n\t\t\tbox-shadow: none;\n\t\t\ttransition: box-shadow .3s ease;\n\t\t}\n\t\t#isw-419a7b82 .isw-card:hover {\n\t\t\tbox-shadow: 0px 14px 36px -8px rgba(99,102,241,0.2);\n\t\t}\n\n\t\t\/* Layer 1 \u2014 Image *\/\n\t\t#isw-419a7b82 .isw-img-wrap {\n\t\t\tposition: absolute; top: 0; left: 0;\n\t\t\twidth: 100%; height: 100%; z-index: 1;\n\t\t}\n\t\t#isw-419a7b82 .isw-img-wrap img {\n\t\t\twidth: 100%; height: 100%; display: block; object-fit: cover;\n\t\t}\n\n\t\t\/* Layer 2 \u2014 Text bar *\/\n\t\t#isw-419a7b82 .isw-text-wrap {\n\t\t\tposition: absolute; bottom: 0; left: 0;\n\t\t\twidth: 100%; height: 35%; z-index: 2;\n\t\t\tbackground: #fff; padding: 16px 24px 20px;\n\t\t\tbox-sizing: border-box; display: flex; flex-direction: column;\n\t\t\tgap: 8px; overflow: hidden;\n\t\t\ttransition: height .35s cubic-bezier(.4,0,.2,1);\n\t\t}\n\t\t#isw-419a7b82 .isw-card:hover .isw-text-wrap { height: 100%; }\n\n\t\t#isw-419a7b82 .isw-excerpt-wrap,\n\t\t#isw-419a7b82 .isw-readmore-wrap { display: none; }\n\t\t#isw-419a7b82 .isw-card:hover .isw-excerpt-wrap { display: block; }\n\t\t#isw-419a7b82 .isw-card:hover .isw-readmore-wrap { display: block; margin-top: auto; }\n\n\t\t\/* \u2500\u2500 Variation 2 overrides \u2500\u2500 *\/\n\t\t#isw-419a7b82.isw-v2 .isw-card {\n\t\t\theight: 530px;\n\t\t\tborder-radius: 15px;\n\t\t\tbox-shadow: none;\n\t\t}\n\t\t#isw-419a7b82.isw-v2 .isw-card:hover {\n\t\t\tbox-shadow: 0px 14px 36px -8px rgba(99,102,241,0.2);\n\t\t}\n\t\t#isw-419a7b82.isw-v2 .isw-img-wrap {\n\t\t\tposition: absolute; top: 0; left: 0;\n\t\t\twidth: 100%; height: 100%; z-index: 1;\n\t\t}\n\t\t#isw-419a7b82.isw-v2 .isw-img-wrap img {\n\t\t\twidth: 100%; height: 100%;\n\t\t\tposition: absolute; top: 0; left: 0;\n\t\t\tobject-fit: cover;\n\t\t}\n\t\t#isw-419a7b82.isw-v2 .isw-text-wrap {\n\t\t\tposition: absolute; bottom: 0; left: 0;\n\t\t\twidth: 100%; height: 120px; z-index: 2;\n\t\t\tpadding: 12px 20px;\n\t\t\tbox-sizing: border-box; display: flex;\n\t\t\tflex-direction: column; gap: 6px;\n\t\t\toverflow: hidden; justify-content: center;\n\t\t\ttransition: height .2s ease-in-out, padding .2s ease-in-out;\n\t\t}\n\t\t#isw-419a7b82.isw-v2 .isw-card:hover .isw-text-wrap {\n\t\t\theight: 100% !important;\n\t\t\tpadding: 24px 20px !important;\n\t\t\tjustify-content: flex-start !important;\n\t\t\toverflow-y: auto !important;\n\t\t}\n\t\t#isw-419a7b82.isw-v2 .isw-title { margin-bottom: 4px !important; margin-top: 0 !important; }\n\t\t#isw-419a7b82.isw-v2 .isw-excerpt { margin: 6px 0 !important; }\n\t\t#isw-419a7b82.isw-v2 .isw-readmore { text-decoration: none; display: inline-block; }\n\t\t#isw-419a7b82.isw-v2 .isw-card:hover .isw-excerpt-wrap { display: block; }\n\t\t#isw-419a7b82.isw-v2 .isw-card:hover .isw-readmore-wrap {\n\t\t\tdisplay: flex; justify-content: flex-end; margin-top: auto;\n\t\t}\n\n\t\t\/* \u2500\u2500 Tag \u2014 no underline, same for V1 and V2 \u2500\u2500 *\/\n\t\t#isw-419a7b82 .isw-tag {\n\t\t\tfont-family: \"Figtree\", sans-serif !important;\n\t\t\tfont-size: 14px !important;\n\t\t\tfont-weight: 600 !important;\n\t\t\ttext-transform: uppercase;\n\t\t\topacity: 0.75;\n\t\t\tmargin: 0;\n\t\t\tline-height: 18px;\n\t\t\tborder-bottom: none !important;\n\t\t\tdisplay: block;\n\t\t}\n\n\t\t\/* \u2500\u2500 Title \u2500\u2500 *\/\n\t\t#isw-419a7b82 .isw-title {\n\t\t\tfont-family: \"Figtree\", sans-serif !important;\n\t\t\tfont-size: 20px !important;\n\t\t\tfont-weight: 700 !important;\n\t\t\tline-height: 24px !important;\n\t\t\tcolor: #171616 !important;\n\t\t\tmargin: 0;\n\t\t\tdisplay: -webkit-box;\n\t\t\t-webkit-line-clamp: 3;\n\t\t\t-webkit-box-orient: vertical;\n\t\t\toverflow: hidden;\n\t\t}\n\t\t#isw-419a7b82 .isw-title a { color: inherit !important; text-decoration: none; }\n\n\t\t\/* \u2500\u2500 Arrows \u2014 outside the slider via widget padding \u2500\u2500 *\/\n\t\t#isw-419a7b82 .isw-prev,\n\t\t#isw-419a7b82 .isw-next {\n\t\t\tposition: absolute;\n\t\t\ttop: calc(50% - 30px); \/* vertically centered on card area *\/\n\t\t\ttransform: translateY(-50%);\n\t\t\tbackground: none !important;\n\t\t\tborder: none !important;\n\t\t\tbox-shadow: none !important;\n\t\t\tpadding: 0;\n\t\t\tcursor: pointer;\n\t\t\twidth: 24px; height: 24px;\n\t\t\tdisplay: flex;\n\t\t\talign-items: center; justify-content: center;\n\t\t\tz-index: 10;\n\t\t\toutline: none;\n\t\t}\n\t\t#isw-419a7b82 .isw-prev { left: 4px; }\n\t\t#isw-419a7b82 .isw-next { right: 4px; }\n\t\t#isw-419a7b82 .isw-prev::after,\n\t\t#isw-419a7b82 .isw-next::after,\n\t\t#isw-419a7b82 .isw-prev::before,\n\t\t#isw-419a7b82 .isw-next::before { display: none !important; content: none !important; }\n\t\t@media(max-width:768px){\n\t\t\t#isw-419a7b82 .isw-prev,\n\t\t\t#isw-419a7b82 .isw-next { display: none !important; }\n\t\t\t#isw-419a7b82 { padding: 0; }\n\t\t\t\/* V2 mobile: fixed image height, card height follows *\/\n\t\t\t#isw-419a7b82.isw-v2 .isw-card { height: 500px !important; }\n\t\t\t#isw-419a7b82.isw-v2 .isw-img-wrap { height: 402px !important; }\n\t\t\t#isw-419a7b82.isw-v2 .isw-img-wrap img { object-fit: fill !important; height: 95% !important; }\n\t\t}\n\n\t\t\/* Dots *\/\n\t\t#isw-419a7b82 .isw-dots {\n\t\t\tdisplay: flex;\n\t\t\talign-items: center;\n\t\t\tjustify-content: center;\n\t\t\tgap: 16px;\n\t\t\tmargin-top: 32px;\n\t\t\theight: 36px;\n\t\t\tbox-sizing: content-box;\n\t\t}\n\t\t#isw-419a7b82 .isw-dot {\n\t\t\twidth: 10px !important;\n\t\t\theight: 10px !important;\n\t\t\tborder-radius: 50% !important;\n\t\t\tbackground: var(--e-global-color-f527530) !important;\n\t\t\topacity: 0.25;\n\t\t\tcursor: pointer;\n\t\t\tborder: none !important;\n\t\t\tpadding: 0;\n\t\t\ttransition: width .25s ease, height .25s ease, opacity .25s ease, background .25s ease, border .25s ease;\n\t\t\tflex-shrink: 0;\n\t\t\tdisplay: block;\n\t\t\tbox-sizing: border-box;\n\t\t}\n\t\t#isw-419a7b82 .isw-dot.active {\n\t\t\twidth: 14px !important;\n\t\t\theight: 14px !important;\n\t\t\topacity: 1 !important;\n\t\t\tbackground: var(--e-global-color-5d532a8) !important;\n\t\t\tborder: 3px solid var(--e-global-color-f527530) !important;\n\t\t}\n\n\t\t\/* Read More *\/\n\t\t#isw-419a7b82 .isw-readmore {\n\t\t\tfont-size: 14px; font-weight: 600;\n\t\t\ttext-decoration: none; color: #171616 !important;\n\t\t\tborder-bottom: 1px solid #171616;\n\t\t\tdisplay: inline-block;\n\t\t}\n\t\t<\/style><div id=\"isw-419a7b82\" class=\"isw-v2\"><button class=\"isw-prev\" aria-label=\"Previous\"><svg width=\"24\" height=\"24\" viewBox=\"0 0 24 24\" fill=\"none\"><path d=\"M15.75 19.5L8.25 12L15.75 4.5\" stroke=\"#000\" stroke-width=\"1.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\"\/><\/svg><\/button><button class=\"isw-next\" aria-label=\"Next\"><svg width=\"24\" height=\"24\" viewBox=\"0 0 24 24\" fill=\"none\"><path d=\"M8.25 4.5L15.75 12L8.25 19.5\" stroke=\"#000\" stroke-width=\"1.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\"\/><\/svg><\/button><div class=\"isw-outer\"><div class=\"isw-track\"><div class=\"isw-slide\"><div class=\"isw-card\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/\" style=\"position:absolute;top:0;left:0;width:100%;height:100%;z-index:1;display:block;text-decoration:none;\"><div class=\"isw-img-wrap\"><img decoding=\"async\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/Securing-AI-Agents-Main-Thumbnail-1024x1024.webp\" alt=\"Securing AI Agents: 4 New Risks and 7 Controls\" loading=\"lazy\"><\/div><\/a><div class=\"isw-text-wrap\"><p class=\"isw-tag\">AI\/ML<\/p><p class=\"isw-title\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/\">Securing AI Agents: 4 New Risks and 7 Controls<\/a><\/p><div class=\"isw-excerpt-wrap\"><p class=\"isw-excerpt\">AI agents act with insider access at machine speed. Why that breaks traditional security models, and seven controls for engineering...<\/p><\/div><div class=\"isw-readmore-wrap\" style=\"text-align:left\"><a class=\"isw-readmore\" href=\"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/\">Read More<\/a><\/div><\/div><\/div><\/div><div class=\"isw-slide\"><div class=\"isw-card\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/corporate-ai-implementation-failure\/\" style=\"position:absolute;top:0;left:0;width:100%;height:100%;z-index:1;display:block;text-decoration:none;\"><div class=\"isw-img-wrap\"><img decoding=\"async\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/Corporate-AI-Implementation-Failure-5-Causes-and-Fixes-Main-Blog-Thumbnail.webp\" alt=\"Corporate AI Implementation Failure: 5 Causes and Fixes\" loading=\"lazy\"><\/div><\/a><div class=\"isw-text-wrap\"><p class=\"isw-tag\">AI\/ML<\/p><p class=\"isw-title\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/corporate-ai-implementation-failure\/\">Corporate AI Implementation Failure: 5 Causes and Fixes<\/a><\/p><div class=\"isw-excerpt-wrap\"><p class=\"isw-excerpt\">Why Corporate AI Implementation Fails &amp; How to Fix It 10Pearls Editorial Team 13 min read Summary Corporate AI implementation...<\/p><\/div><div class=\"isw-readmore-wrap\" style=\"text-align:left\"><a class=\"isw-readmore\" href=\"https:\/\/10pearls.com\/uk\/blog\/corporate-ai-implementation-failure\/\">Read More<\/a><\/div><\/div><\/div><\/div><div class=\"isw-slide\"><div class=\"isw-card\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/how-to-choose-ai-development-partner\/\" style=\"position:absolute;top:0;left:0;width:100%;height:100%;z-index:1;display:block;text-decoration:none;\"><div class=\"isw-img-wrap\"><img decoding=\"async\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/An-enterprise-guide-for-choosing-an-AI-development-partner-thumbnail.webp\" alt=\"How to Choose an AI Development Partner: Red Flags to Avoid\" loading=\"lazy\"><\/div><\/a><div class=\"isw-text-wrap\"><p class=\"isw-tag\">AI\/ML<\/p><p class=\"isw-title\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/how-to-choose-ai-development-partner\/\">How to Choose an AI Development Partner: Red Flags to Avoid<\/a><\/p><div class=\"isw-excerpt-wrap\"><p class=\"isw-excerpt\">An AI project can impress in a demo and still fall apart on real data and real users. How to...<\/p><\/div><div class=\"isw-readmore-wrap\" style=\"text-align:left\"><a class=\"isw-readmore\" href=\"https:\/\/10pearls.com\/uk\/blog\/how-to-choose-ai-development-partner\/\">Read More<\/a><\/div><\/div><\/div><\/div><div class=\"isw-slide\"><div class=\"isw-card\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/llms-in-healthcare-use-cases-risks-deployment\/\" style=\"position:absolute;top:0;left:0;width:100%;height:100%;z-index:1;display:block;text-decoration:none;\"><div class=\"isw-img-wrap\"><img decoding=\"async\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/llms-healthcare-use-cases-risks-thumbnail.webp\" alt=\"LLM in Healthcare: Use Cases, Risks &amp; Deployment\" loading=\"lazy\"><\/div><\/a><div class=\"isw-text-wrap\"><p class=\"isw-tag\">AI\/ML<\/p><p class=\"isw-title\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/llms-in-healthcare-use-cases-risks-deployment\/\">LLM in Healthcare: Use Cases, Risks &amp; Deployment<\/a><\/p><div class=\"isw-excerpt-wrap\"><p class=\"isw-excerpt\">A guide to LLM in healthcare, covering use cases, model selection, risks, HIPAA compliance, and what organisations need for deployment.<\/p><\/div><div class=\"isw-readmore-wrap\" style=\"text-align:left\"><a class=\"isw-readmore\" href=\"https:\/\/10pearls.com\/uk\/blog\/llms-in-healthcare-use-cases-risks-deployment\/\">Read More<\/a><\/div><\/div><\/div><\/div><div class=\"isw-slide\"><div class=\"isw-card\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/ai-governance-checklist-9-controls-to-enforce\/\" style=\"position:absolute;top:0;left:0;width:100%;height:100%;z-index:1;display:block;text-decoration:none;\"><div class=\"isw-img-wrap\"><img decoding=\"async\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/AI-Governance-and-Compliance-\u2014-A-Checklist-for-Enterprise-AI-Deployment-Main-Blog-Thumbnail.webp\" alt=\"AI Governance Checklist: 9 Controls to Enforce\" loading=\"lazy\"><\/div><\/a><div class=\"isw-text-wrap\"><p class=\"isw-tag\">AI\/ML<\/p><p class=\"isw-title\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/ai-governance-checklist-9-controls-to-enforce\/\">AI Governance Checklist: 9 Controls to Enforce<\/a><\/p><div class=\"isw-excerpt-wrap\"><p class=\"isw-excerpt\">AI now writes a large share of production code, and it does not just change how software is built, it...<\/p><\/div><div class=\"isw-readmore-wrap\" style=\"text-align:left\"><a class=\"isw-readmore\" href=\"https:\/\/10pearls.com\/uk\/blog\/ai-governance-checklist-9-controls-to-enforce\/\">Read More<\/a><\/div><\/div><\/div><\/div><div class=\"isw-slide\"><div class=\"isw-card\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/how-ai-is-changing-healthcare-cybersecurity\/\" style=\"position:absolute;top:0;left:0;width:100%;height:100%;z-index:1;display:block;text-decoration:none;\"><div class=\"isw-img-wrap\"><img decoding=\"async\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/How-AI-Is-Reshaping-Healthcare-Cybersecurity-Threats-Main-Blog-Thumbnail-1024x1024.webp\" alt=\"How AI Is Reshaping Healthcare Cybersecurity Threats\" loading=\"lazy\"><\/div><\/a><div class=\"isw-text-wrap\"><p class=\"isw-tag\">AI\/ML<\/p><p class=\"isw-title\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/how-ai-is-changing-healthcare-cybersecurity\/\">How AI Is Reshaping Healthcare Cybersecurity Threats<\/a><\/p><div class=\"isw-excerpt-wrap\"><p class=\"isw-excerpt\">AI is helping both sides of healthcare cybersecurity. Attackers use it to speed up reconnaissance, write convincing phishing and deploy...<\/p><\/div><div class=\"isw-readmore-wrap\" style=\"text-align:left\"><a class=\"isw-readmore\" href=\"https:\/\/10pearls.com\/uk\/blog\/how-ai-is-changing-healthcare-cybersecurity\/\">Read More<\/a><\/div><\/div><\/div><\/div><div class=\"isw-slide\"><div class=\"isw-card\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/ai-for-credit-unions\/\" style=\"position:absolute;top:0;left:0;width:100%;height:100%;z-index:1;display:block;text-decoration:none;\"><div class=\"isw-img-wrap\"><img decoding=\"async\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/AI-for-Credit-Unions-Use-Cases-Agents-and-ROI-Main-Blog-Thumbnail-1024x1024.webp\" alt=\"AI for Credit Unions: Use Cases, Agents and ROI\" loading=\"lazy\"><\/div><\/a><div class=\"isw-text-wrap\"><p class=\"isw-tag\">AI\/ML<\/p><p class=\"isw-title\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/ai-for-credit-unions\/\">AI for Credit Unions: Use Cases, Agents and ROI<\/a><\/p><div class=\"isw-excerpt-wrap\"><p class=\"isw-excerpt\">A practical guide to AI ROI at institutions under $1B, covering high-payback use cases, real integration costs, NCUA expectations, and...<\/p><\/div><div class=\"isw-readmore-wrap\" style=\"text-align:left\"><a class=\"isw-readmore\" href=\"https:\/\/10pearls.com\/uk\/blog\/ai-for-credit-unions\/\">Read More<\/a><\/div><\/div><\/div><\/div><div class=\"isw-slide\"><div class=\"isw-card\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/aml-transaction-monitoring\/\" style=\"position:absolute;top:0;left:0;width:100%;height:100%;z-index:1;display:block;text-decoration:none;\"><div class=\"isw-img-wrap\"><img decoding=\"async\" src=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/aml-transaction-monitoring-thumbnail.webp\" alt=\"AML Transaction Monitoring and KYC: From Manual to AI\" loading=\"lazy\"><\/div><\/a><div class=\"isw-text-wrap\"><p class=\"isw-tag\">AI\/ML<\/p><p class=\"isw-title\"><a href=\"https:\/\/10pearls.com\/uk\/blog\/aml-transaction-monitoring\/\">AML Transaction Monitoring and KYC: From Manual to AI<\/a><\/p><div class=\"isw-excerpt-wrap\"><p class=\"isw-excerpt\">Traditional KYC and AML processes are struggling to keep up with modern financial crime. Learn how automation and AI are...<\/p><\/div><div class=\"isw-readmore-wrap\" style=\"text-align:left\"><a class=\"isw-readmore\" href=\"https:\/\/10pearls.com\/uk\/blog\/aml-transaction-monitoring\/\">Read More<\/a><\/div><\/div><\/div><\/div><\/div><\/div><div class=\"isw-dots\"><\/div><\/div><script>(function(){\n\t\t\tvar id       = \"isw-419a7b82\";\n\t\t\tvar spvDesk  = 3;\n\t\t\tvar spvTab   = 2;\n\t\t\tvar gap      = 20;\n\t\t\tvar total    = 8;\n\t\t\tvar autoplay = true;\n\t\t\tvar autoSpd  = 5000;\n\t\t\tvar showDots = true;\n\n\t\t\tvar wrap  = document.getElementById(id);\n\t\t\tif(!wrap) return;\n\t\t\tvar outer  = wrap.querySelector(\".isw-outer\");\n\t\t\tvar track  = wrap.querySelector(\".isw-track\");\n\t\t\tvar slides = wrap.querySelectorAll(\".isw-slide\");\n\t\t\tvar dotsWrap = wrap.querySelector(\".isw-dots\");\n\t\t\tvar btnPrev  = wrap.querySelector(\".isw-prev\");\n\t\t\tvar btnNext  = wrap.querySelector(\".isw-next\");\n\n\t\t\tvar cur      = 0;\n\t\t\tvar timer    = null;\n\t\t\tvar spv      = spvDesk;\n\n\t\t\tfunction getSpv(){\n\t\t\t\tvar w = window.innerWidth;\n\t\t\t\tif(w <= 480)  return 1;\n\t\t\t\tif(w <= 900)  return spvTab;\n\t\t\t\treturn spvDesk;\n\t\t\t}\n\n\t\t\tfunction setSlideWidths(){\n\t\t\t\tspv = getSpv();\n\t\t\t\tvar outerW = outer.offsetWidth;\n\t\t\t\tvar slideW = (outerW - gap * (spv - 1)) \/ spv;\n\t\t\t\t[].forEach.call(slides, function(s){\n\t\t\t\t\ts.style.width    = slideW + \"px\";\n\t\t\t\t\ts.style.minWidth = slideW + \"px\";\n\t\t\t\t});\n\t\t\t}\n\n\t\t\tfunction maxPage(){ return Math.max(0, total - spv); }\n\n\t\t\tfunction goTo(n, instant){\n\t\t\t\tn = Math.max(0, Math.min(n, maxPage()));\n\t\t\t\tcur = n;\n\t\t\t\tsetSlideWidths();\n\t\t\t\tvar slideW = slides[0] ? slides[0].offsetWidth : 0;\n\t\t\t\tvar offset = cur * (slideW + gap);\n\t\t\t\ttrack.style.transition = instant ? \"none\" : \"transform .5s cubic-bezier(.25,.46,.45,.94)\";\n\t\t\t\ttrack.style.transform  = \"translateX(-\" + offset + \"px)\";\n\t\t\t\tupdDots();\n\t\t\t}\n\n\t\t\tfunction updDots(){\n\t\t\t\tif(!showDots || !dotsWrap) return;\n\t\t\t\tvar pages = maxPage() + 1;\n\t\t\t\t\/\/ Rebuild dots if count changed\n\t\t\t\tif(dotsWrap.children.length !== pages){\n\t\t\t\t\tdotsWrap.innerHTML = \"\";\n\t\t\t\t\tfor(var i=0;i<pages;i++){\n\t\t\t\t\t\t(function(idx){\n\t\t\t\t\t\t\tvar b = document.createElement(\"button\");\n\t\t\t\t\t\t\tb.className = \"isw-dot\" + (idx===0?\" active\":\"\");\n\t\t\t\t\t\t\tb.setAttribute(\"aria-label\",\"Page \"+(idx+1));\n\t\t\t\t\t\t\tb.addEventListener(\"click\",function(){ stopTimer(); goTo(idx); startTimer(); });\n\t\t\t\t\t\t\tdotsWrap.appendChild(b);\n\t\t\t\t\t\t})(i);\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t\t[].forEach.call(dotsWrap.querySelectorAll(\".isw-dot\"),function(d,i){\n\t\t\t\t\td.classList.toggle(\"active\", i===cur);\n\t\t\t\t});\n\t\t\t}\n\n\t\t\tfunction stopTimer(){ clearInterval(timer); timer=null; }\n\t\t\tfunction startTimer(){\n\t\t\t\tif(!autoplay) return; stopTimer();\n\t\t\t\ttimer = setInterval(function(){\n\t\t\t\t\tgoTo(cur >= maxPage() ? 0 : cur + 1);\n\t\t\t\t}, autoSpd);\n\t\t\t}\n\n\t\t\tif(btnPrev) btnPrev.addEventListener(\"click\",function(){ stopTimer(); goTo(cur-1); startTimer(); });\n\t\t\tif(btnNext) btnNext.addEventListener(\"click\",function(){ stopTimer(); goTo(cur+1); startTimer(); });\n\n\t\t\t\/\/ Touch swipe\n\t\t\tvar tx=0, ty=0;\n\t\t\touter.addEventListener(\"touchstart\",function(e){ tx=e.touches[0].clientX; ty=e.touches[0].clientY; stopTimer(); },{passive:true});\n\t\t\touter.addEventListener(\"touchend\",function(e){\n\t\t\t\tvar dx=e.changedTouches[0].clientX-tx, dy=e.changedTouches[0].clientY-ty;\n\t\t\t\tif(Math.abs(dx)>Math.abs(dy)&&Math.abs(dx)>40){ goTo(dx<0?cur+1:cur-1); }\n\t\t\t\tstartTimer();\n\t\t\t},{passive:true});\n\n\t\t\t\/\/ Resize\n\t\t\tvar rt;\n\t\t\twindow.addEventListener(\"resize\",function(){\n\t\t\t\tclearTimeout(rt);\n\t\t\t\trt=setTimeout(function(){ goTo(cur,true); },200);\n\t\t\t});\n\n\t\t\t\/\/ Init \u2014 use requestAnimationFrame so layout is painted and offsetWidth is correct\n\t\t\tfunction doInit(){\n\t\t\t\tsetSlideWidths();\n\t\t\t\tupdDots();\n\t\t\t\tgoTo(0, true);\n\t\t\t\ttrack.style.visibility = \"visible\"; \/\/ show after widths are set\n\t\t\t\tstartTimer();\n\t\t\t}\n\t\t\tif(document.readyState === \"loading\"){\n\t\t\t\tdocument.addEventListener(\"DOMContentLoaded\", function(){ requestAnimationFrame(doInit); });\n\t\t\t} else {\n\t\t\t\trequestAnimationFrame(doInit);\n\t\t\t}\n\t\t})();<\/script>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>AI agents act with insider access at machine speed. Why that breaks traditional security models, and seven controls for engineering agents securely.<\/p>\n","protected":false},"author":87,"featured_media":222864,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[49],"tags":[4585],"class_list":["post-222862","post","type-post","status-publish","format-standard","has-post-thumbnail","category-blog","tag-ai-and-ml"],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v25.5 (Yoast SEO v28.4) - https:\/\/yoast.com\/product\/yoast-seo-premium-wordpress\/ -->\n<title>Securing AI Agents: 4 New Risks and 7 Controls<\/title>\n<meta name=\"description\" content=\"Securing AI agents takes more than traditional controls. How agentic AI security risks differ, and seven controls that keep agents in bounds.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/\" \/>\n<meta property=\"og:locale\" content=\"en_GB\" \/>\n<meta property=\"og:type\" content=\"website\" \/>\n<meta property=\"og:title\" content=\"Securing AI Agents: 4 New Risks and 7 Controls\" \/>\n<meta property=\"og:description\" content=\"Securing AI agents takes more than traditional controls. How agentic AI security risks differ, and seven controls that keep agents in bounds.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/\" \/>\n<meta property=\"og:site_name\" content=\"10Pearls UK\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/10Pearls\/\" \/>\n<meta property=\"article:published_time\" content=\"2026-11-04T05:39:54+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-11-04T07:59:20+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/Securing-AI-Agents-LinkThumbnail.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"640\" \/>\n\t<meta property=\"og:image:height\" content=\"359\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Muhammad Yaqoob\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:title\" content=\"Securing AI Agents: 4 New Risks and 7 Controls\" \/>\n<meta name=\"twitter:description\" content=\"Securing AI agents takes more than traditional controls. How agentic AI security risks differ, and seven controls that keep agents in bounds.\" \/>\n<meta name=\"twitter:creator\" content=\"@TenPearls\" \/>\n<meta name=\"twitter:site\" content=\"@TenPearls\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Muhammad Yaqoob\" \/>\n\t<meta name=\"twitter:label2\" content=\"Estimated reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"10 minutes\" \/>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Securing AI Agents: 4 New Risks and 7 Controls","description":"Securing AI agents takes more than traditional controls. How agentic AI security risks differ, and seven controls that keep agents in bounds.","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:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/","og_locale":"en_GB","og_type":"website","og_title":"Securing AI Agents: 4 New Risks and 7 Controls","og_description":"Securing AI agents takes more than traditional controls. How agentic AI security risks differ, and seven controls that keep agents in bounds.","og_url":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/","og_site_name":"10Pearls UK","article_publisher":"https:\/\/www.facebook.com\/10Pearls\/","article_published_time":"2026-11-04T05:39:54+00:00","article_modified_time":"2026-11-04T07:59:20+00:00","og_image":[{"width":640,"height":359,"url":"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/Securing-AI-Agents-LinkThumbnail.webp","type":"image\/webp"}],"author":"Muhammad Yaqoob","twitter_card":"summary_large_image","twitter_title":"Securing AI Agents: 4 New Risks and 7 Controls","twitter_description":"Securing AI agents takes more than traditional controls. How agentic AI security risks differ, and seven controls that keep agents in bounds.","twitter_creator":"@TenPearls","twitter_site":"@TenPearls","twitter_misc":{"Written by":"Muhammad Yaqoob","Estimated reading time":"10 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/#article","isPartOf":{"@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/"},"author":{"name":"Muhammad Yaqoob","@id":"https:\/\/10pearls.com\/uk\/#\/schema\/person\/20e745b0085af58ce70c776e8e4e7bbb"},"headline":"Securing AI Agents: 4 New Risks and 7 Controls","datePublished":"2026-11-04T05:39:54+00:00","dateModified":"2026-11-04T07:59:20+00:00","mainEntityOfPage":{"@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/"},"wordCount":2113,"publisher":{"@id":"https:\/\/10pearls.com\/uk\/#organization"},"image":{"@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/#primaryimage"},"thumbnailUrl":"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/Securing-AI-Agents-Main-Thumbnail.webp","keywords":["AI\/ML"],"articleSection":["Blog"],"inLanguage":"en-GB"},{"@type":"WebPage","@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/","url":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/","name":"Securing AI Agents: 4 New Risks and 7 Controls","isPartOf":{"@id":"https:\/\/10pearls.com\/uk\/#website"},"primaryImageOfPage":{"@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/#primaryimage"},"image":{"@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/#primaryimage"},"thumbnailUrl":"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/Securing-AI-Agents-Main-Thumbnail.webp","datePublished":"2026-11-04T05:39:54+00:00","dateModified":"2026-11-04T07:59:20+00:00","description":"Securing AI agents takes more than traditional controls. How agentic AI security risks differ, and seven controls that keep agents in bounds.","breadcrumb":{"@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/#breadcrumb"},"inLanguage":"en-GB","potentialAction":[{"@type":"ReadAction","target":["https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/"]}]},{"@type":"ImageObject","inLanguage":"en-GB","@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/#primaryimage","url":"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/Securing-AI-Agents-Main-Thumbnail.webp","contentUrl":"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2026\/10\/Securing-AI-Agents-Main-Thumbnail.webp","width":4096,"height":4096},{"@type":"BreadcrumbList","@id":"https:\/\/10pearls.com\/uk\/blog\/securing-ai-agents\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/10pearls.com\/uk\/"},{"@type":"ListItem","position":2,"name":"Securing AI Agents: 4 New Risks and 7 Controls"}]},{"@type":"WebSite","@id":"https:\/\/10pearls.com\/uk\/#website","url":"https:\/\/10pearls.com\/uk\/","name":"10Pearls UK","description":"AI-Powered Software Development Company","publisher":{"@id":"https:\/\/10pearls.com\/uk\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/10pearls.com\/uk\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-GB"},{"@type":"Organization","@id":"https:\/\/10pearls.com\/uk\/#organization","name":"10Pearls","alternateName":"Software development company","url":"https:\/\/10pearls.com\/uk\/","logo":{"@type":"ImageObject","inLanguage":"en-GB","@id":"https:\/\/10pearls.com\/uk\/#\/schema\/logo\/image\/","url":"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2025\/07\/10Pearls-Black.png","contentUrl":"https:\/\/10pearls.com\/uk\/wp-content\/uploads\/2025\/07\/10Pearls-Black.png","width":282,"height":107,"caption":"10Pearls"},"image":{"@id":"https:\/\/10pearls.com\/uk\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/10Pearls\/","https:\/\/x.com\/TenPearls","https:\/\/www.linkedin.com\/company\/10pearls\/"],"description":"10Pearls is a leading AI-powered software development company. We build intelligent digital products for healthcare, finance, retail, and more.","email":"info@10pearls.com","telephone":"+44 7957 620104","legalName":"10Pearls"},{"@type":"Person","@id":"https:\/\/10pearls.com\/uk\/#\/schema\/person\/20e745b0085af58ce70c776e8e4e7bbb","name":"Muhammad Yaqoob","image":{"@type":"ImageObject","inLanguage":"en-GB","@id":"https:\/\/secure.gravatar.com\/avatar\/c3b7cdd5d81b799dcca57c38f657bad931be08e8ec4c9429fde1e39caa5f11eb?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/c3b7cdd5d81b799dcca57c38f657bad931be08e8ec4c9429fde1e39caa5f11eb?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/c3b7cdd5d81b799dcca57c38f657bad931be08e8ec4c9429fde1e39caa5f11eb?s=96&d=mm&r=g","caption":"Muhammad Yaqoob"}}]}},"_links":{"self":[{"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/posts\/222862","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/users\/87"}],"replies":[{"embeddable":true,"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/comments?post=222862"}],"version-history":[{"count":10,"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/posts\/222862\/revisions"}],"predecessor-version":[{"id":223607,"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/posts\/222862\/revisions\/223607"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/media\/222864"}],"wp:attachment":[{"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/media?parent=222862"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/categories?post=222862"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/10pearls.com\/uk\/wp-json\/wp\/v2\/tags?post=222862"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}