{"id":4233,"date":"2025-04-10T09:27:38","date_gmt":"2025-04-10T13:27:38","guid":{"rendered":"https:\/\/www.indigoconsulting.ca\/?p=4233"},"modified":"2025-10-08T00:51:59","modified_gmt":"2025-10-08T04:51:59","slug":"securing-agentic-ai-with-authentication-authorization-standards","status":"publish","type":"post","link":"https:\/\/www.indigoconsulting.ca\/fr\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/","title":{"rendered":"The Missing Link: Securing Agentic AI Interactions with Authentication &#038; Authorization Standards"},"content":{"rendered":"<div data-elementor-type=\"wp-post\" data-elementor-id=\"4233\" class=\"elementor elementor-4233\" data-elementor-post-type=\"post\">\n\t\t\t\t<div class=\"elementor-element elementor-element-b438f28 e-flex e-con-boxed e-con e-parent\" data-id=\"b438f28\" 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-339e212 e-con-full e-flex e-con e-child\" data-id=\"339e212\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t<div class=\"elementor-element elementor-element-c20927e e-con-full e-flex e-con e-child\" data-id=\"c20927e\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t<div class=\"elementor-element elementor-element-04eb2d3 elementor-widget elementor-widget-text-editor\" data-id=\"04eb2d3\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><a href=\"https:\/\/www.indigoconsulting.ca\/fr\/\"><strong>Accueil<\/strong><\/a><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-0cc2000 elementor-view-default elementor-widget elementor-widget-icon\" data-id=\"0cc2000\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\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<i aria-hidden=\"true\" class=\"fas fa-chevron-right\"><\/i>\t\t\t<\/div>\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-114d2c7 elementor-widget elementor-widget-text-editor\" data-id=\"114d2c7\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><a href=\"https:\/\/www.indigoconsulting.ca\/fr\/solutions\/resources\/\">Ressources<\/a><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-8a404c1 elementor-view-default elementor-widget elementor-widget-icon\" data-id=\"8a404c1\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"icon.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\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<i aria-hidden=\"true\" class=\"fas fa-chevron-right\"><\/i>\t\t\t<\/div>\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-c04279b elementor-widget elementor-widget-text-editor\" data-id=\"c04279b\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p>The Missing Link<\/p>\t\t\t\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\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-3c34fb4 e-flex e-con-boxed e-con e-parent\" data-id=\"3c34fb4\" 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-d27403f elementor-widget elementor-widget-heading\" data-id=\"d27403f\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">Securing Agentic AI Interactions with Authentication &amp; Authorization Standards<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-080ac04 elementor-widget elementor-widget-text-editor\" data-id=\"080ac04\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p>Large language models (LLMs) are highly effective at processing information independently, but they face challenges when tasks demand knowledge beyond their current training datasets.<\/p><p>For AI Agents to be truly useful to the users, they need timely access to relevant context, such as file stores, knowledge bases, real time data feeds, or enterprise datastores, and develop the ability to take actions like updating documents or writing emails. Nowadays, integrating AI with these sources has been a complex process. It often requires developers to write custom code or rely on specialized plugins for each data source or API, resulting in fragile, hard-to-scale systems. The current lack of adaptability and versatility in agents makes it difficult to streamline the integration of external functionalities. This limitation prevents agents from performing complex tasks across a wide range of data sources. For example, if you wanted an AI agent to access both a file server and a database, you would need to integrate both the file server&#8217;s API and a database driver. Each of these integrations has its own authentication, data format, and potential issues, and is closely tied to custom code, connectors, or plugins for a specific data source.<\/p><p>In a word, the current state is fragmented as many frameworks exist and standards are lacking, which leads to multiple challenges in particular for the security and compliance of handling authentication, access control, and real-time monitoring at scale.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6194434 elementor-widget elementor-widget-text-editor\" data-id=\"6194434\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<h2><strong>\u00a0Agentic AI Considerations<\/strong><\/h2>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6fe1066 elementor-widget elementor-widget-text-editor\" data-id=\"6fe1066\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<ul><li aria-level=\"1\"><b>Trust and Privacy Concerns:<\/b> How can we establish trust in AI agents and ensure that they operate on behalf of THE human user? How can we safeguard the privacy of enterprise users and prevent the exposure of sensitive information, such as denying the AI inquiries to other employee\u2019s salary details?<\/li><li aria-level=\"1\"><b>Authorization Granularity Concern:<\/b> Can we move beyond coarse-grained permissions, such as OAuth scopes,\u00a0 to provide the necessary fine-grained controls for agents (what, when, where, and how)?<\/li><li aria-level=\"1\"><b>Accountability Concern:<\/b> How can we effectively audit agent actions and attribute them to the authorizing human user and agent, especially when issues arise?<\/li><li aria-level=\"1\"><b>Consent Control Concern<\/b>: Could we establish mechanisms that allow users to dynamically provide and revoke consent, adapting permissions as agents operate?<\/li><li aria-level=\"1\"><b>Frictionless Experience Concern:<\/b> How can we create a frictionless authentication and authorization process, especially for consumers? Ensure a balance between safety and frictionless user experience.<br \/><br \/><\/li><\/ul><p>This white paper delves into a framework for secure delegation to AI agents, addressing the critical challenges of authorization, accountability, and access control in the rapidly expanding field of Agentic AI. By extending established internet authentication protocols (OAuth 2.0 and OpenID Connect) with AI-specific integrations, we can ensure compatibility with existing systems while enabling granular control over AI agent capabilities.<\/p><p>Recently, Anthropic introduced the Model Context Protocol (MCP), an open standard designed to connect AI agents with various data and tool sources. At its core, MCP has the potential to empower AI agents with greater autonomy, structured modularity, and reusable cognitive modules, enhancing both efficiency and trustworthiness. This enables AI to operate in a more human-like manner while also scaling seamlessly for multi-agent collaboration. This potential of MCP to transform the development of agentic AI systems has been recognized by major AI players.<\/p>\t\t\t\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-f9aaba4 e-flex e-con-boxed e-con e-parent\" data-id=\"f9aaba4\" 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-01110ce elementor-widget elementor-widget-text-editor\" data-id=\"01110ce\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<h2>\u00a0<\/h2><h2><strong>Impersonation or Delegation?<\/strong><\/h2>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6f0b009 elementor-widget elementor-widget-text-editor\" data-id=\"6f0b009\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><b>Impersonation<\/b> is an approach where an AI Agent assumes fully the user\u2019s identity, making it indistinguishable from an actual user when performing actions. The AI agent logs in as a user, either using an impersonation token or the user credentials. The agent will have all the permissions that a user does, and would be treated exactly as a user would.<\/p><p>With <b>delegation<\/b>, the AI Agent is granted specific, limited permissions, to act on behalf of the user without assuming full user identity. The user explicitly grants permissions to an AI agent to perform certain specific actions within predefined boundaries, typically implemented using OAuth 2.0 delegated authorization, access token or fine-grained authorization.<\/p><p>While impersonation may be easier to implement for common scenarios, delegated access tokens offer superior control and security. With impersonation, there are significant security concerns around controlling access levels. <i>Delegated access allows users to define the AI\u2019s permissions precisely, aligning better with existing standards and offering greater future-proofing.<\/i><\/p><p><i><b>Which one to use?<\/b><\/i><\/p><p>The challenge with impersonation is that it\u2019s difficult to control the level of access granted. While a claim can be included to indicate that an AI agent is using the token, and the token itself is specifically for the AI agent, impersonation risks bypassing that safeguard. This could lead to the AI accessing areas it shouldn\u2019t.<\/p><p>With a delegation workflow, a user can define the AI\u2019s access, limiting it to specific tasks. The AI can then work behind the scenes on those tasks only. Impersonation lacks this control, potentially requiring additional logic and safeguards \u2013 essentially reinventing what already exists in other workflows.<\/p>\t\t\t\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-eab5d17 e-flex e-con-boxed e-con e-parent\" data-id=\"eab5d17\" 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-88ba2dd elementor-widget elementor-widget-text-editor\" data-id=\"88ba2dd\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<h2><b><span style=\"color: #003b59;\">Delegation Token U<\/span><span style=\"color: #003b59;\">sing the MCP Pattern Example<\/span><\/b><\/h2>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-5fdd79a elementor-widget elementor-widget-image\" data-id=\"5fdd79a\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img fetchpriority=\"high\" decoding=\"async\" width=\"768\" height=\"638\" src=\"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/MCPCLIENTAPP-768x638.png\" class=\"attachment-medium_large size-medium_large wp-image-4548\" alt=\"MCPCLIENTAPP\" srcset=\"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/MCPCLIENTAPP-768x638.png 768w, https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/MCPCLIENTAPP-300x249.png 300w, https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/MCPCLIENTAPP-1024x851.png 1024w, https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/MCPCLIENTAPP-1536x1276.png 1536w, https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/MCPCLIENTAPP-2048x1702.png 2048w\" sizes=\"(max-width: 768px) 100vw, 768px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-56d19bc elementor-widget elementor-widget-text-editor\" data-id=\"56d19bc\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><br \/>So how would a simple example work when trying to incorporate both delegated tokens and an MCP pattern?<\/p><p>The MCP architecture consists of an MCP client, which can be seen as a native app to an authentication server, it interacts with LLM Provider and MCP server, a flow will look like:<\/p><ol><li>The user also\u0192 authenticates and receives a user token.<\/li><li>The user&#8217;s query is sent to the client<\/li><li>A token exchange flow can then be used to get an actor token for MCP client\/native app.<\/li><li>MCP Client utilizes the LLM provider and interacts with the MCP server.<\/li><li>The MCP server then calls the API to the application server<\/li><li>API or the Web App (registered with the authentication server) requests the Authentication server to validate the user ID token and actor token.<\/li><li>The Web application API retrieves information based on the user&#8217;s authorization scope (assume only scope is used for now, but more advanced AuthZ can be used), then it returns the information back to the requesting MCP client that can call the LLM provider to respond to the user\u2019s query.<\/li><\/ol><p><br \/>Artificial intelligence, particularly in the form of sophisticated AI Agents, is rapidly transforming how we interact with technology. These agents promise unprecedented efficiency by accessing data and performing actions on our behalf. However, realizing this potential hinges on solving a critical challenge:<em> securely granting AI agents the access they need without compromising user data, privacy, or enterprise security.<\/em><\/p><p>As highlighted, integrating these agents with the diverse systems they need to interact with (databases, APIs, file stores) is currently a fragmented, complex process lacking standardization.<\/p><p>Let us now focus on the authentication and authorization (AuthN\/AuthZ) piece of this puzzle \u2013 how we verify the agent&#8217;s right to act and control what it can do, especially when acting for a user. We&#8217;ll explore the specific security and control problems this creates, examine current solutions using existing standards, and look towards emerging protocols designed to build a more secure and trustworthy AI ecosystem.<\/p>\t\t\t\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-644c4d6 e-flex e-con-boxed e-con e-parent\" data-id=\"644c4d6\" 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-2005f53 elementor-widget elementor-widget-text-editor\" data-id=\"2005f53\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<h2><b>&nbsp;<\/b><\/h2>\n<h2><b>The Problem Deep Dive<br>Fragmentation, Control, and Trust<\/b><\/h2>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-f387f9e elementor-widget elementor-widget-image\" data-id=\"f387f9e\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1536\" height=\"1017\" src=\"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/AIAugFragmnt-1536x1017.png\" class=\"attachment-1536x1536 size-1536x1536 wp-image-4549\" alt=\"AI Augmented Fragmentation Illustration\" srcset=\"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/AIAugFragmnt-1536x1017.png 1536w, https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/AIAugFragmnt-300x199.png 300w, https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/AIAugFragmnt-1024x678.png 1024w, https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/AIAugFragmnt-768x509.png 768w, https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/10\/AIAugFragmnt-2048x1356.png 2048w\" sizes=\"(max-width: 1536px) 100vw, 1536px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-78e3210 elementor-widget elementor-widget-text-editor\" data-id=\"78e3210\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p>Integrating AI securely faces significant hurdles stemming from the lack of standardized approaches, directly impacting the key concerns outlined earlier:<\/p><ol><li aria-level=\"1\"><b>API Design &amp; Integration Complexity:<\/b> <br \/>Most existing APIs weren&#8217;t built for autonomous AI agents needing delegated authority. Integrating requires bespoke solutions for each data source, leading to fragile systems that are hard to scale and secure consistently, directly impacting the goal of a Frictionless Experience.<\/li><li aria-level=\"1\"><b>Maintaining User Control &amp; Consent:<\/b> <br \/>Users <i>must<\/i> remain in control. How do we build systems where consent is clear, granular, dynamic, and easily revocable (Consent Control Concern)? Ad-hoc methods erode trust.<\/li><li aria-level=\"1\"><b>Trust &amp; Privacy Risks:<\/b> <br \/>Without clear standards for delegation, how do we ensure an agent acts <i>only<\/i> as intended by the user (Trust Concern)? In enterprise settings, preventing inadvertent access to sensitive data like salaries or confidential documents is paramount (Privacy Concern).<\/li><li aria-level=\"1\"><b>Insufficient Authorization Granularity:<\/b> Traditional permissions, like broad OAuth scopes (e.g., read_files), often lack the nuance needed for AI agents. We need finer control \u2013 perhaps allowing reading <i>specific types<\/i> of files but not others, or only performing actions within certain contexts (Authorization Granularity Concern).<\/li><li aria-level=\"1\"><b>Accountability Gaps:<\/b> <br \/>If an AI agent performs an incorrect or malicious action, how do we trace it back? We need clear audit trails differentiating actions taken directly by the user versus those taken by an agent on their behalf, attributing responsibility correctly (Accountability Concern).<\/li><li aria-level=\"1\"><b>Past Efforts &amp; Stalled Standards:<\/b> <br \/>While protocols like OAuth 2.0 Token Exchange (RFC 8693) provide a foundation for exchanging tokens, potentially for delegation, they\u00a0 haven&#8217;t been universally adopted or extended with specific profiles for the complex needs of AI agents.<\/li><li aria-level=\"1\"><b>Divergent Needs &#8211; Enterprise vs. Consumer<\/b>:<ul><li aria-level=\"2\"><b>Enterprise:<\/b> Focuses heavily on strict data governance, addressing Trust, Privacy<b>, <\/b>Granularity<b>,<\/b> and Accountability.<\/li><li aria-level=\"2\"><b>Consumer:<\/b> Prioritizes a Frictionless Experience while still needing robust underlying security and Consent Control.<\/li><\/ul><\/li><\/ol><p>\u00a0<\/p>\t\t\t\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-9bdc6c1 e-flex e-con-boxed e-con e-parent\" data-id=\"9bdc6c1\" 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-1bf978e elementor-widget elementor-widget-text-editor\" data-id=\"1bf978e\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<h2><b>Current Solutions: <br>Extending OAuth for Secure Delegation<\/b><\/h2>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-837b1d3 elementor-widget elementor-widget-text-editor\" data-id=\"837b1d3\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p>In the absence of definitive AI-specific standards, several patterns have emerged, primarily leveraging and extending OAuth 2.0:<\/p><h6><b style=\"font-size: 1.25rem; font-family: inherit; text-align: var(--text-align);\">Table 1: Comparison of Token-Based Authentication Approaches for AI Agents<\/b><\/h6>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-4e8de7e pp-table-center elementor-widget elementor-widget-pp-table\" data-id=\"4e8de7e\" data-element_type=\"widget\" data-e-type=\"widget\" data-settings=\"{&quot;table_type&quot;:&quot;responsive&quot;,&quot;scrollable&quot;:&quot;no&quot;}\" data-widget_type=\"pp-table.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<div class=\"pp-table-container\">\n\t\t\t<table \n\t\t\tclass=\"pp-table tablesaw\" data-tablesaw-mode=\"stack\"\t\t\t\t>\n\t\t\t\t\t\t\t<colgroup>\n\t\t\t\t\t\t\t\t<col\n\t\t\t\t\tspan=\"1\"\n\t\t\t\t\tclass=\"elementor-repeater-item-59e1fb8\"\n\t\t\t\t\t\t\t\t\t\t\tstyle=\"\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\"\n\t\t\t\t\t\t\t\t\t\t>\n\t\t\t\t\t\t\t<\/colgroup>\n\t\t\t\t\t\t<thead>\n\t\t\t\t<tr class=\"pp-table-row\">\n\t\t\t\t\t<th class=\"pp-table-cell pp-table-cell-678ec53\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Approach<\/span><\/span><\/th><th class=\"pp-table-cell pp-table-cell-be8233d\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Description<\/span><\/span><\/th><th class=\"pp-table-cell pp-table-cell-939f8a9\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Advantages<\/span><\/span><\/th><th class=\"pp-table-cell pp-table-cell-c28a80a\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Disadvantages<\/span><\/span><\/th>\t\t\t\t<\/tr>\n\t\t\t<\/thead>\n\t\t\t\t\t<tfoot>\n\t\t\t<\/tr>\t\t<\/tfoot>\n\t\t\t\t<tbody>\n\t\t\t<tr ><td class=\"pp-table-cell pp-table-cell-9c90eac\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Trust the AI<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-1ce6c55\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">No specific authentication mechanism. Sharing credentials.<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-a467978\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Simple to implement (initially).<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-c22e1a9\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">High security risk, no accountability, potential for misuse.<\/span><\/span><\/td><\/tr><tr ><td class=\"pp-table-cell pp-table-cell-048effc\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Share User Access Token<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-7a10a23\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Providing the AI agent with the user\u2019s existing access token.<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-06a54b4\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Allows the agent to act with the user\u2019s full permissions.<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-460346b\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">High security risk if the agent is compromised, violates least privilege.<\/span><\/span><\/td><\/tr><tr ><td class=\"pp-table-cell pp-table-cell-bfc9af8\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Impersonation Token (User)<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-102a211\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">AI acts on a user\u2019s token with an identifier for the agent (sub=user, act=agent).<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-66b7985\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Provides auditability of agent actions, operates within user\u2019s permissions.<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-fbc386e\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Still relies on the user\u2019s token being securely managed.<\/span><\/span><\/td><\/tr><tr ><td class=\"pp-table-cell pp-table-cell-268b771\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Delegation Token(Agent)<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-3da9086\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">AI has its own token, acts on behalf of the user (\u201con_behalf_of\u201d claim).<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-8a12ac7\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">Better separation of concerns, potentially more granular control over agent permissions, enhanced auditability.<\/span><\/span><\/td><td class=\"pp-table-cell pp-table-cell-fdba686\"><span class=\"pp-table-cell-content\"><span class=\"pp-table-cell-text\">More complex implementation than simply sharing user tokens.<\/span><\/span><\/td><\/tr>\t\t<\/tbody>\n\t\t\t\t\t<\/table>\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-b3f11b8 elementor-widget elementor-widget-text-editor\" data-id=\"b3f11b8\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><b>Worst Solution<br \/><\/b><b>Blind Trust &amp; Brute Force:<\/b> Relying on screen scraping or simulating user input is insecure, unreliable, and provides zero accountability. <b>Avoid.<\/b><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-b47edd7 elementor-widget-divider--view-line elementor-widget elementor-widget-divider\" data-id=\"b47edd7\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"divider.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<div class=\"elementor-divider\">\n\t\t\t<span class=\"elementor-divider-separator\">\n\t\t\t\t\t\t<\/span>\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-e254334 elementor-widget elementor-widget-text-editor\" data-id=\"e254334\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><b>Bad Solution<br \/><\/b><b>Sharing the User\u2019s Access Token:<\/b> Giving the AI the user\u2019s primary token grants excessive permissions and makes auditing impossible (violates <b>Accountability<\/b> and <b>Granularity<\/b> principles).<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-3728cb9 elementor-widget-divider--view-line elementor-widget elementor-widget-divider\" data-id=\"3728cb9\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"divider.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<div class=\"elementor-divider\">\n\t\t\t<span class=\"elementor-divider-separator\">\n\t\t\t\t\t\t<\/span>\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-68a958e elementor-widget elementor-widget-text-editor\" data-id=\"68a958e\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><b>Workable Solution<br \/><\/b><b>Impersonation Token (AI <\/b><b><i>as<\/i><\/b><b> User):<\/b> Using <a href=\"https:\/\/www.rfc-editor.org\/rfc\/rfc8693.html\">OAuth 2.0 Token Exchange (RFC 8693)<\/a>, the AI exchanges appropriate credentials for a <i>new<\/i>token that identifies the user (sub claim) but includes an indicator that an agent is acting (e.g., act claim). This offers <i>some<\/i> <b>Accountability<\/b> by showing delegation occurred. Several IAM vendors support variations (though this list is not exhaustive):<\/p><ul><li aria-level=\"2\"><b>PingAM:<\/b><a href=\"https:\/\/docs.pingidentity.com\/pingam\/7.5\/oauth2-guide\/token-exchange-impersonation.html\"> Token Exchange Impersonation Docs<\/a><\/li><li aria-level=\"2\"><b>Curity:<\/b><a href=\"https:\/\/curity.io\/resources\/learn\/impersonation-implementation\/\"> Impersonation Implementation<\/a>,<a href=\"https:\/\/curity.io\/resources\/learn\/impersonation-flow-approaches\/\"> Flow Approaches<\/a><\/li><li aria-level=\"2\"><b>WSO2 IS:<\/b><a href=\"https:\/\/is.docs.wso2.com\/en\/next\/guides\/authorization\/user-impersonation\/\"> User Impersonation Guide<\/a><\/li><li aria-level=\"2\"><b>Zitadel:<\/b><a href=\"https:\/\/zitadel.com\/docs\/guides\/integrate\/token-exchange\"> Token Exchange Guide<\/a><\/li><\/ul>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1bbe328 elementor-widget-divider--view-line elementor-widget elementor-widget-divider\" data-id=\"1bbe328\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"divider.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<div class=\"elementor-divider\">\n\t\t\t<span class=\"elementor-divider-separator\">\n\t\t\t\t\t\t<\/span>\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-2bfee9b elementor-widget elementor-widget-text-editor\" data-id=\"2bfee9b\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><b>Better Solution<br \/>Token Delegation (AI <\/b><b><i>on behalf of<\/i><\/b><b> User):<\/b> A more robust pattern where the AI obtains a token identifying <i>itself<\/i> (sub = AI_Agent_ID) while clearly indicating it acts <i>on behalf of<\/i> the user (e.g., on_behalf_of=user_ID or structured act claim). This provides superior Accountability (clear agent identity) and enables better Authorization Granularity (permissions can be tailored to the agent acting for that user). This approach best reflects a secure delegation model.<\/p>\t\t\t\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-9aba505 e-flex e-con-boxed e-con e-parent\" data-id=\"9aba505\" 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-5d25314 elementor-widget elementor-widget-text-editor\" data-id=\"5d25314\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<h2>\u00a0<\/h2><h2><strong>Did Someone Mention Authorization?<\/strong><\/h2>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-a9afe2d elementor-widget elementor-widget-text-editor\" data-id=\"a9afe2d\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p>The intersection of JWT-based service authorization and agentic AI creates a unique category of complex problems.<\/p><p>The \u201ceasy\u201d way to handle authorization is through scopes directly in the token. That actually does make it quite simple to validate the token, and see if they are allowed to do what they say they do:<\/p><ul><li aria-level=\"1\">From an API standpoint: This Token is issued by an Authorization Server (AS) I trust, the signature is good, the Audience is me, and I even recognize the impersonation claims!<\/li><li aria-level=\"1\">Do the Scopes match the URI path policy? Yes? Let\u2019s go!<br \/><br \/><\/li><\/ul><p>But oftentimes it\u2019s not so simple. Without devolving into a <a href=\"https:\/\/www.indigoconsulting.ca\/fr\/authorization-demystified-your-gateway-to-secure-access\/\">Discussion about Authorization Policies<\/a> &amp; scope creep, authorization policies gain a lot of flexibility and power when attribute and relationship based aspects of the user are taken into account.<\/p><p>Lets say you were trying to make a fine grained policy decision for an AI working on behalf of a human manager (the subject) at a company:<\/p><ol><li aria-level=\"1\">The AI should only be aware of employees and perform actions only on employees where the subject is listed as the manager<\/li><li aria-level=\"1\">This AI might have further restrictions that a human might not have:<ul><li aria-level=\"1\">Authorization to modify employee status, but NOT to delete.<\/li><li>Exporting of data that would have to be cleansed of PII.<\/li><li>Visibility to only a subset of user information etc\u2026<\/li><\/ul><\/li><li>All actions might furthermore be subject to additional approvals after the fact, especially depending on the context of what is being done! (though this is more of an Identity Governance problem than an authorization problem, but one leads to another!)<br \/><br \/><\/li><\/ol><p>What is important to realize is: that modern and advanced Authorization tools are already powerful enough to deal with Agentic AI problems. However, they do need to have custom policies crafted in order to take into account the specifics of AI enabled work-loads, and enforce the perimeter of what the AI can and cannot do within a given API or service.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-fb42329 elementor-widget-divider--view-line elementor-widget elementor-widget-divider\" data-id=\"fb42329\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"divider.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<div class=\"elementor-divider\">\n\t\t\t<span class=\"elementor-divider-separator\">\n\t\t\t\t\t\t<\/span>\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-df3986c elementor-widget elementor-widget-text-editor\" data-id=\"df3986c\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<h2>\u00a0<strong>Emerging Protocols &amp; Future Standards<\/strong><\/h2>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-43cb911 elementor-widget elementor-widget-text-editor\" data-id=\"43cb911\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p>While extending OAuth provides workable solutions <i>now<\/i>, the industry is developing next-generation standards better suited for AI:<\/p><ol><li aria-level=\"1\"><b>Model Context Protocol (MCP):<\/b> Introduced by Anthropic in November 2024 (<a href=\"https:\/\/modelcontextprotocol.io\/\">https:\/\/modelcontextprotocol.io\/<\/a>), MCP is an open standard specifically designed to structure how AI models securely connect to and receive context from external data sources and tools. This enables AI to operate in a more human-like manner while also scaling seamlessly for multi-agent collaboration. It aims to standardize the <i>input<\/i> to the AI, including authentication\/authorization information potentially derived from tokens obtained via the patterns above. Its goal is to address the integration fragmentation and provide a common language, potentially improving <b>Trust<\/b> and enabling easier implementation of <b>Granularity<\/b> and <b>Consent<\/b>. (See<a href=\"https:\/\/modelcontextprotocol.io\/development\/roadmap\"> Roadmap<\/a>).<\/li><li aria-level=\"1\"><b>GNAP (Grant Negotiation and Authorization Protocol):<\/b> As <a href=\"https:\/\/www.rfc-editor.org\/rfc\/rfc9635.html\">RFC 9635, GNAP<\/a> aims to modernize and simplify authorization, potentially offering more flexible ways to handle the complex delegation scenarios needed for AI agents, improving upon OAuth 2.0\u2019s limitations.<\/li><li aria-level=\"1\"><b>UCAN (User-Controlled Authorization Network):<\/b> This specification (<a href=\"https:\/\/github.com\/ucan-wg\/spec\">https:\/\/github.com\/ucan-wg\/spec<\/a>) focuses on decentralized, user-controlled permissions using cryptographic capabilities (\u201ccapabilities-based security\u201d). This directly addresses <b>Consent Control<\/b> and <b>Granularity<\/b>by allowing users to delegate very specific, verifiable permissions.<\/li><\/ol><p>\u00a0<\/p>\t\t\t\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-15bc44f e-flex e-con-boxed e-con e-parent\" data-id=\"15bc44f\" 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-789a50d elementor-widget elementor-widget-text-editor\" data-id=\"789a50d\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<h2>\u00a0<\/h2><h2><strong>Conclusion &amp; Recommendations<\/strong><\/h2>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-4b8858c elementor-widget elementor-widget-text-editor\" data-id=\"4b8858c\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Securely integrating AI Agents requires moving beyond ad-hoc methods towards standardized, robust authentication and authorization frameworks. Addressing concerns around trust, privacy, granularity, accountability, consent, and user experience is paramount.<\/span><\/p><h5><strong>OUR RECOMMENDATIONS<\/strong><\/h5><ol><li style=\"font-weight: 400;\" aria-level=\"1\"><b>Prioritize Secure Delegation Patterns:<\/b><span style=\"font-weight: 400;\"> Adopt the &#8220;Better Solution&#8221; (Token Delegation via OAuth) where possible, or at minimum the &#8220;Workable Solution&#8221; (Token Impersonation), leveraging RFC 8693 and capabilities within your IAM platform. Avoid insecure shortcuts.<\/span><\/li><li style=\"font-weight: 400;\" aria-level=\"1\"><b>Focus on Granularity and Auditability:<\/b><span style=\"font-weight: 400;\"> Design systems that allow for fine-grained permissions specific to AI agent tasks and ensure all actions are logged with clear attribution to both the user and the agent.<\/span><\/li><li style=\"font-weight: 400;\" aria-level=\"1\"><b>Implement Clear Consent Mechanisms:<\/b><span style=\"font-weight: 400;\"> Ensure users have transparent and manageable ways to grant and revoke agent permissions.<\/span><\/li><li style=\"font-weight: 400;\" aria-level=\"1\"><b>Explore Emerging Standards:<\/b><span style=\"font-weight: 400;\"> Stay informed about MCP, GNAP, and UCAN. Participate in discussions and consider how these standards can solve deeper integration and authorization challenges. Practical exploration, such as building proofs-of-concept integrating MCP with robust AuthN\/AuthZ patterns, will be key to understanding their real-world application.<\/span><\/li><\/ol><div>\u00a0<\/div><p><span style=\"font-weight: 400;\">Building a future where AI agents are powerful, helpful, <i>and<\/i> trustworthy requires a concerted effort focused on secure foundations. By extending proven standards today and embracing promising new protocols tomorrow, we can bridge the missing link in AI security.<\/span><\/p>\t\t\t\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-5a803e9 e-flex e-con-boxed e-con e-parent\" data-id=\"5a803e9\" 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-968cf28 elementor-widget elementor-widget-text-editor\" data-id=\"968cf28\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<section class=\"elementor-element elementor-element-edit-mode elementor-element-b0d2849 elementor-section elementor-top-section elementor-section-boxed elementor-section-height-default\" data-id=\"b0d2849\" data-element_type=\"section\" data-model-cid=\"c51\"><div class=\"elementor-container elementor-column-gap-default ui-sortable ui-sortable-disabled\"><div class=\"elementor-element elementor-element-edit-mode elementor-element-95025db elementor-element--toggle-edit-tools elementor-column elementor-top-column\" data-id=\"95025db\" data-element_type=\"column\" data-model-cid=\"c52\" data-col=\"100\"><div class=\"elementor-widget-wrap elementor-element-populated ui-sortable ui-sortable-disabled\"><div class=\"elementor-element elementor-element-edit-mode elementor-element-a0dac6a elementor-element--toggle-edit-tools elementor-widget elementor-widget-heading\" data-id=\"a0dac6a\" data-element_type=\"widget\" data-model-cid=\"c53\" data-widget_type=\"heading.default\"><div class=\"elementor-widget-container\"><p class=\"elementor-heading-title elementor-size-default elementor-inline-editing\" data-elementor-setting-key=\"title\"><strong>Authors:<\/strong> <br \/>Aditya Sharma, <br \/>Boo Leong Khoo, <br \/>Dominik Ludera, <br \/>Grace Zhang, <br \/>Wyatt Bourdeau, <br \/>Nicolas Seigneur and <br \/>Paul Figura<\/p><\/div><\/div><\/div><\/div><\/div><\/section>\t\t\t\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-a5d929c e-flex e-con-boxed e-con e-parent\" data-id=\"a5d929c\" 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-d9511f7 e-con-full e-flex e-con e-child\" data-id=\"d9511f7\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t<div class=\"elementor-element elementor-element-0c45f64 e-con-full e-flex e-con e-child\" data-id=\"0c45f64\" data-element_type=\"container\" data-e-type=\"container\" data-settings=\"{&quot;background_background&quot;:&quot;classic&quot;}\">\n\t\t<div class=\"elementor-element elementor-element-a2f67fc e-con-full e-flex e-con e-child\" data-id=\"a2f67fc\" data-element_type=\"container\" data-e-type=\"container\">\n\t\t\t\t<div class=\"elementor-element elementor-element-b78ea56 elementor-widget elementor-widget-heading\" data-id=\"b78ea56\" data-element_type=\"widget\" data-e-type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">Pr\u00eat \u00e0 simplifier<br> la gestion de l'IAM<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-cbe66f8 elementor-align-center elementor-invisible elementor-widget elementor-widget-button\" data-id=\"cbe66f8\" data-element_type=\"widget\" data-e-type=\"widget\" data-settings=\"{&quot;_animation&quot;:&quot;fadeIn&quot;,&quot;_animation_delay&quot;:200}\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm\" href=\"https:\/\/www.indigoconsulting.ca\/fr\/contact\/\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t\t\t\t<span class=\"elementor-button-text\">Prendre rendez-vous<\/span>\n\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t\t\t<\/div>\n\t\t\t\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>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>","protected":false},"excerpt":{"rendered":"<p>Home Resources The Missing Link Securing Agentic AI Interactions with Authentication &amp; Authorization Standards Large language models (LLMs) are highly effective at processing information independently, but they face challenges when tasks demand knowledge beyond their current training datasets. For AI Agents to be truly useful to the users, they need timely access to relevant context, [&hellip;]<\/p>\n","protected":false},"author":21,"featured_media":4236,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"content-type":"","footnotes":""},"categories":[117],"tags":[],"class_list":["post-4233","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-blog"],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v25.8 (Yoast SEO v28.2) - https:\/\/yoast.com\/product\/yoast-seo-premium-wordpress\/ -->\n<title>Securing Agentic AI: How to Use OAuth &amp; MCP for Trusted Authentication and Authorization<\/title>\n<meta name=\"description\" content=\"Discover how to securely empower AI agents with delegated access using standards like OAuth 2.0, OpenID Connect, and Model Context Protocol (MCP). Learn how to address trust, privacy, and fine-grained authorization challenges in Agentic AI systems.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.indigoconsulting.ca\/fr\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/\" \/>\n<meta property=\"og:locale\" content=\"fr_CA\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"The Missing Link: Securing Agentic AI Interactions with Authentication &amp; Authorization Standards\" \/>\n<meta property=\"og:description\" content=\"Discover how to securely empower AI agents with delegated access using standards like OAuth 2.0, OpenID Connect, and Model Context Protocol (MCP). Learn how to address trust, privacy, and fine-grained authorization challenges in Agentic AI systems.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.indigoconsulting.ca\/fr\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/\" \/>\n<meta property=\"og:site_name\" content=\"Indigo Consulting\" \/>\n<meta property=\"article:published_time\" content=\"2025-04-10T13:27:38+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2025-10-08T04:51:59+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/04\/ai-blog.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1080\" \/>\n\t<meta property=\"og:image:height\" content=\"1080\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Web Master\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Web Master\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"12 minutes\" \/>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Securing Agentic AI: How to Use OAuth & MCP for Trusted Authentication and Authorization","description":"Discover how to securely empower AI agents with delegated access using standards like OAuth 2.0, OpenID Connect, and Model Context Protocol (MCP). Learn how to address trust, privacy, and fine-grained authorization challenges in Agentic AI systems.","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:\/\/www.indigoconsulting.ca\/fr\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/","og_locale":"fr_CA","og_type":"article","og_title":"The Missing Link: Securing Agentic AI Interactions with Authentication & Authorization Standards","og_description":"Discover how to securely empower AI agents with delegated access using standards like OAuth 2.0, OpenID Connect, and Model Context Protocol (MCP). Learn how to address trust, privacy, and fine-grained authorization challenges in Agentic AI systems.","og_url":"https:\/\/www.indigoconsulting.ca\/fr\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/","og_site_name":"Indigo Consulting","article_published_time":"2025-04-10T13:27:38+00:00","article_modified_time":"2025-10-08T04:51:59+00:00","og_image":[{"width":1080,"height":1080,"url":"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/04\/ai-blog.png","type":"image\/png"}],"author":"Web Master","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Web Master","Est. reading time":"12 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/#article","isPartOf":{"@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/"},"author":{"name":"Web Master","@id":"https:\/\/www.indigoconsulting.ca\/#\/schema\/person\/38f543d7b3a9f166761affde4e68fb3f"},"headline":"The Missing Link: Securing Agentic AI Interactions with Authentication &#038; Authorization Standards","datePublished":"2025-04-10T13:27:38+00:00","dateModified":"2025-10-08T04:51:59+00:00","mainEntityOfPage":{"@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/"},"wordCount":2602,"publisher":{"@id":"https:\/\/www.indigoconsulting.ca\/#organization"},"image":{"@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/#primaryimage"},"thumbnailUrl":"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/04\/ai-blog.png","articleSection":["Blog"],"inLanguage":"fr-CA"},{"@type":"WebPage","@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/","url":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/","name":"Securing Agentic AI: How to Use OAuth & MCP for Trusted Authentication and Authorization","isPartOf":{"@id":"https:\/\/www.indigoconsulting.ca\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/#primaryimage"},"image":{"@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/#primaryimage"},"thumbnailUrl":"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/04\/ai-blog.png","datePublished":"2025-04-10T13:27:38+00:00","dateModified":"2025-10-08T04:51:59+00:00","description":"Discover how to securely empower AI agents with delegated access using standards like OAuth 2.0, OpenID Connect, and Model Context Protocol (MCP). Learn how to address trust, privacy, and fine-grained authorization challenges in Agentic AI systems.","breadcrumb":{"@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/#breadcrumb"},"inLanguage":"fr-CA","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/"]}]},{"@type":"ImageObject","inLanguage":"fr-CA","@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/#primaryimage","url":"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/04\/ai-blog.png","contentUrl":"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2025\/04\/ai-blog.png","width":1080,"height":1080},{"@type":"BreadcrumbList","@id":"https:\/\/www.indigoconsulting.ca\/blog\/securing-agentic-ai-with-authentication-authorization-standards\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.indigoconsulting.ca\/"},{"@type":"ListItem","position":2,"name":"The Missing Link: Securing Agentic AI Interactions with Authentication &#038; Authorization Standards"}]},{"@type":"WebSite","@id":"https:\/\/www.indigoconsulting.ca\/#website","url":"https:\/\/www.indigoconsulting.ca\/","name":"Indigo Consulting","description":"A Leading IAM, Compliance, &amp; IT Consultant","publisher":{"@id":"https:\/\/www.indigoconsulting.ca\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.indigoconsulting.ca\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"fr-CA"},{"@type":"Organization","@id":"https:\/\/www.indigoconsulting.ca\/#organization","name":"Indigo Consulting","url":"https:\/\/www.indigoconsulting.ca\/","logo":{"@type":"ImageObject","inLanguage":"fr-CA","@id":"https:\/\/www.indigoconsulting.ca\/#\/schema\/logo\/image\/","url":"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2020\/03\/logo_indigo.png","contentUrl":"https:\/\/www.indigoconsulting.ca\/wp-content\/uploads\/2020\/03\/logo_indigo.png","width":363,"height":109,"caption":"Indigo Consulting"},"image":{"@id":"https:\/\/www.indigoconsulting.ca\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.linkedin.com\/company\/indigo-technologies-canada-inc.\/"]},{"@type":"Person","@id":"https:\/\/www.indigoconsulting.ca\/#\/schema\/person\/38f543d7b3a9f166761affde4e68fb3f","name":"Web Master","image":{"@type":"ImageObject","inLanguage":"fr-CA","@id":"https:\/\/secure.gravatar.com\/avatar\/545c64cbccfbdd8a03700444efeb701f8d8efbce05186f326dae06e6d17e5575?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/545c64cbccfbdd8a03700444efeb701f8d8efbce05186f326dae06e6d17e5575?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/545c64cbccfbdd8a03700444efeb701f8d8efbce05186f326dae06e6d17e5575?s=96&d=mm&r=g","caption":"Web Master"},"url":"https:\/\/www.indigoconsulting.ca\/fr\/author\/webmaster\/"}]}},"_links":{"self":[{"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/posts\/4233","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/users\/21"}],"replies":[{"embeddable":true,"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/comments?post=4233"}],"version-history":[{"count":0,"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/posts\/4233\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/media\/4236"}],"wp:attachment":[{"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/media?parent=4233"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/categories?post=4233"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.indigoconsulting.ca\/fr\/wp-json\/wp\/v2\/tags?post=4233"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}