BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//cfp.cloud-native.rejekts.io//ZCZSMW
BEGIN:VTIMEZONE
TZID:EST
BEGIN:STANDARD
DTSTART:20001029T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10;UNTIL=20061029T070000Z
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
END:STANDARD
BEGIN:STANDARD
DTSTART:20071104T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000402T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4;UNTIL=20060402T080000Z
TZNAME:EDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20070311T030000
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:EDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-cloud-native-rejekts-atlanta-na-atlanta-2025-ZCZSMW@cfp.cloud-n
 ative.rejekts.io
DTSTART;TZID=EST:20251108T111000
DTEND;TZID=EST:20251108T114000
DESCRIPTION:# The Paranoid's Guide to Deploying Skynet's Interns\n\nSo\, yo
 u've built an AI Agent. Congratulations! It's brilliant\, autonomous\, and
  probably a little bit terrifying. While we're all racing to build the nex
 t generation of intelligent applications\, we're bolting them onto deploym
 ent architectures that treat them like any other legacy system or worse\, 
 blindly deploying them without a plan. This is a mistake\, and it's going 
 to get weird.\n\nThis talk presents a reference deployment architecture fo
 r AI Agent applications\, starting with a quick primer on their core compo
 nents: the **Agents**\, the **MCP servers**\, the **Tools** they access\, 
 and the **Memory** that gives them context. Then\, we dive into the deep e
 nd of the security nightmare they represent.\n\nWe'll explore the messy re
 ality of modern AI deployments:\n\n- **A Tangled Web of Trust:** Agents an
 d MCPs are exposed to a chaotic mix of tools and services with wildly diff
 erent levels of trust. How do you keep your high-security internal tool fr
 om being manipulated by an agent that just scraped a questionable Reddit t
 hread?\n\n- **Persistent Threats:** The very nature of an Agent's memory m
 eans that attacks and threats can persist and evolve across sessions. A vu
 lnerability exploited today could be a weapon wielded by the agent tomorro
 w.\n\n- **Amplified Supply Chain Risks:** Autonomous AI actions turn opaqu
 e\, previously inaccessible components into active parts of your supply ch
 ain. This dramatically increases the attack surface\, making vulnerabiliti
 es that were once theoretical suddenly very exploitable.\n\n- **Compoundin
 g Complexity:** The introduction of multi-agent communication protocols an
 d centralized MCP servers adds layers of complexity that can obscure risk 
 and reduce control when you need it most.\n\nThe core of this talk is a si
 mple\, radical recommendation: **true\, paranoid\, and unapologetic isolat
 ion at every level of the AI Agent application stack.** We'll argue that A
 I components are dynamic\, untrusted supply chains and must be handled wit
 h the same (if not more) scrutiny as any other production system.\n\nYou w
 ill leave this session understanding why segmentation of components by tru
 st level isn't just a good idea\, but absolutely vital. We'll show you why
  you need *more* control over your MCP servers\, not less\, and provide a 
 practical\, defense-in-depth architecture for deploying AI Agents that won
 't turn on you.
DTSTAMP:20261013T125512Z
LOCATION:Crystal Dining Room
SUMMARY:The Paranoid's Guide to Deploying Skynet's Interns - Dan Fernandez
URL:https://cfp.cloud-native.rejekts.io/cloud-native-rejekts-atlanta-na-atl
 anta-2025/talk/ZCZSMW/
END:VEVENT
END:VCALENDAR