BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Chicago BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0600 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:CDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0500 TZOFFSETTO:-0600 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:CST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260805T155005Z UID:237E1C06-F1B6-49A9-BD08-7D8B204A31E8 DTSTART;TZID=America/Chicago:20260921T193000 DTEND;TZID=America/Chicago:20260921T210000 DESCRIPTION:Abstract:\n\nHow Identity\, Enterprise Context\, Observability\ , and Bounded Authority Shape Live Risk Decisions\n\nAn aircraft may begin its journey with a verified pilot\, an approved flight plan\, and permiss ion to take off. But that approval does not guarantee that the original ro ute will remain safe for the entire flight. Weather\, traffic\, fuel\, air port availability\, and other conditions can change. Air traffic control u ses live signals to decide whether the aircraft should continue\, slow dow n\, hold\, change its route\, divert\, or land.\n\nAI agents face a simila r challenge. An agent may authenticate\, receive permissions\, access appr oved tools\, and begin a task correctly. But as it works\, its behavior\, enterprise context\, risk\, cost\, data sensitivity\, and operating condit ions can change. A permission that was appropriate at the start of the ses sion may not be enough to safely approve every action that follows.\n\nThi s talk explains how organizations can build a runtime posture for AI agent s. Runtime posture brings together signals such as agent identity\, the us er it represents\, session state\, business purpose\, enterprise context\, tool usage\, data sensitivity\, behavior\, approvals\, risk\, and cost.\n \nObservability provides live information about what the agent is doing. E nterprise context explains the business meaning\, relationships\, ownershi p\, dependencies\, policies\, and current conditions around an action. Bou nded authority defines the limits within which the agent may operate\, inc luding its allowed tools\, data\, purpose\, time\, risk\, and cost.\n\nUsi ng a practical enterprise scenario and the air traffic control analogy\, t he session will show how these signals can be evaluated at important decis ion points. Based on the agent’s current posture\, the governing system can allow an action\, apply additional limits\, request human approval\, r edirect the agent to a safer option\, or block the action.\n\nThe key mess age is simple: initial authorization starts the journey\, but runtime post ure determines whether the next action is still safe.\n\nVirtual: https:// events.vtools.ieee.org/m/571392 LOCATION:Virtual: https://events.vtools.ieee.org/m/571392 ORGANIZER:wluo@stmarytx.edu SEQUENCE:9 SUMMARY:From Signals to Decisions: Building Runtime Posture for AI Agents URL;VALUE=URI:https://events.vtools.ieee.org/m/571392 X-ALT-DESC:Description: <br /><p><strong>Abstract:</strong></p>\n<p><strong data-olk-copy-source="MessageBody">How Identity\, Enterprise Context\, Ob servability\, and Bounded Authority Shape Live Risk Decisions</strong></p> \n<p>An aircraft may begin its journey with a verified pilot\, an approved flight plan\, and permission to take off. But that approval does not guar antee that the original route will remain safe for the entire flight. Weat her\, traffic\, fuel\, airport availability\, and other conditions can cha nge. Air traffic control uses live signals to decide whether the aircraft should continue\, slow down\, hold\, change its route\, divert\, or land.< /p>\n<p>AI agents face a similar challenge. An agent may authenticate\, re ceive permissions\, access approved tools\, and begin a task correctly. Bu t as it works\, its behavior\, enterprise context\, risk\, cost\, data sen sitivity\, and operating conditions can change. A permission that was appr opriate at the start of the session may not be enough to safely approve ev ery action that follows.</p>\n<p>This talk explains how organizations can build a runtime posture for AI agents. Runtime posture brings together sig nals such as agent identity\, the user it represents\, session state\, bus iness purpose\, enterprise context\, tool usage\, data sensitivity\, behav ior\, approvals\, risk\, and cost.</p>\n<p>Observability provides live inf ormation about what the agent is doing. Enterprise context explains the bu siness meaning\, relationships\, ownership\, dependencies\, policies\, and current conditions around an action. Bounded authority defines the limits within which the agent may operate\, including its allowed tools\, data\, purpose\, time\, risk\, and cost.</p>\n<p>Using a practical enterprise sc enario and the air traffic control analogy\, the session will show how the se signals can be evaluated at important decision points. Based on the age nt&rsquo\;s current posture\, the governing system can allow an action\, a pply additional limits\, request human approval\, redirect the agent to a safer option\, or block the action.</p>\n<p>The key message is simple: ini tial authorization starts the journey\, but runtime posture determines whe ther the next action is still safe.</p> END:VEVENT END:VCALENDAR