BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//bsidesluxembourg-2026//talk//GLKSMY
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-bsidesluxembourg-2026-GLKSMY@pretalx.com
DTSTART;TZID=CET:20260507T133000
DTEND;TZID=CET:20260507T140000
DESCRIPTION:AI agents are no longer simple chatbots—they're autonomous sy
 stems equipped with powerful tools including shell access\, file operation
 s\, and database queries. But what happens when an attacker asks nicely?\n
 \nIn this talk\, we present a real-world vulnerability discovered in a pro
 duction AI platform where we achieved full system command execution throug
 h natural language conversation. Starting with simple reconnaissance. When
  the AI initially denied access\, we researched and deployed a jailbreak t
 echnique that bypassed safety guardrails—all through conversation.\n\nTh
 e result? Reading /etc/passwd\, enumerating system information\, and letti
 ng the AI run reconnaissance commands for us. No credentials. No exploits.
  Just conversation.\n\nAttendees will learn:\n- How AI agent architectures
  create new attack surfaces\n- Practical jailbreak techniques for tool-ena
 bled LLMs\n- The "Confused Deputy" problem in AI systems\n- Defense strate
 gies for securing AI agents
DTSTAMP:20260412T024819Z
LOCATION:IFEN room 2\, Workshops and AI Security Village  (Building D)
SUMMARY:Talk to a Shell : Exploiting AI agent in Real Time - Parth Shukla
URL:https://pretalx.com/bsidesluxembourg-2026/talk/GLKSMY/
END:VEVENT
END:VCALENDAR