Data Security Decoded – "Agentic AI and Identity Sprawl"

Host: Caleb Tolan
Guest: Joe Hladic, Head of Rubrik Zero Labs
Release Date: November 18, 2025

Episode Overview

This episode dives deep into identity-driven threats in modern enterprises, with special focus on the challenges posed by the explosive growth of non-human identities (including AI agents), and the realities of cyber resilience and recovery. Host Caleb Tolan is joined by Joe Hladic of Rubrik Zero Labs, whose team recently released the "Identity Crisis: Understanding and Building Resilience Against Identity Driven Threats" report. The conversation covers the latest threat vectors, practical recovery strategies, the fast adoption (and misconceptions) around agentic AI, and how organizations should prioritize defending across a sprawling landscape of identities.

Key Discussion Points and Insights

The Evolving Identity Landscape

• Identity Is the New Perimeter
  Joe reiterates the industry shift: hybrid infrastructures have dissolved the traditional network perimeter; identity has replaced it as the core boundary to protect.

  "The perimeter is no longer the network, identity is the new perimeter." (03:36, Joe Hladic)

• Non-human Identities’ Rapid Growth
  Non-human identities (service accounts, API keys, AI agents, etc.) now outnumber human ones by roughly 82 to 1. This creates a massive, hard-to-monitor attack surface that traditional security controls aren't built to manage.

  "For every human identity, there are 82 non-human identities." (04:14, Joe Hladic)

• Diverse and Unmonitored Identity Types
  Each type of identity (humans, APIs, AI agents) requires a different approach for effective monitoring and defense, adding complexity and risk.

Prioritizing Recovery and Resilience

• Prioritization Is Business-Specific
  Recovery and resilience strategies must be tailored to an organization’s unique context, critical assets, and dependencies—there’s no one-size-fits-all approach.

  "What is the quickest way to achieve or get back to continuity? ... You have to understand the organization first." (06:24, Joe Hladic)

• Dependency and Asset Mapping are Crucial
  Knowing the minimal viable components to restore (especially for core business functions such as e-commerce websites) is vital for a swift recovery.

  "Have a really solid asset mapping tool and map out your dependencies." (07:44, Joe Hladic)

• Recovery Times are Getting Longer
  Survey data shows shrinking confidence in fast recovery:

  • Only 28% believe they could fully recover in 12 hours or less, down from 43% in 2024.
  • 89% of victims paid a ransom to recover data or stop an attack—highlighting both desperation and the critical nature of prepared recovery plans.

  "Businesses are paying the ransoms 89% of the time... because they're not proactively doing this." (09:54, Joe Hladic)

• Resilience is a Cross-Functional Challenge
  True resilience requires coordination across IT, legal, communications, and security—teams that may rarely collaborate under normal circumstances.

  "Resilience is a combination of a lot of different components and teams... that a lot of times in normal business operations they're not used to doing." (10:54, Joe Hladic)

• Backups are Now Prime Targets
  Attackers increasingly aim for backup systems, knowing that destroying or encrypting backup data puts ultimate leverage over the business.

  "If you own the backups, you own the business." (12:17, Joe Hladic)

The Agentic AI Surge: Reality vs. Hype

• Survey: Nearly 90% Using AI Agents
  Caleb notes that 89% of respondents claim partial or full deployment of agentic AI in their identity stack—a shockingly high number, possibly inflated.

• True Adoption Likely Overstated
  Joe is skeptical: widespread use likely reflects early experimentation in operations, sales, or engineering—not full-scale, security-critical adoption.

  "From a security perspective, I don't really see it happening... Right now it's like proof of concept." (14:38, Joe Hladic)

• Threat Actor Adoption: Not Imminent (Yet)
  Sophisticated nation-state-level attackers, who value operational control and stealth, won’t rush to employ AI agents unless there’s proven upside.

  "AI still hallucinate[s]. Humans, in my opinion, are still a lot more trustworthy." (15:37, Joe Hladic)
  "If you're performing espionage... the whole point is to be quiet and subtle... That's why I'm hesitant that agentic AI is really going to be leveraged in that space." (20:00, Joe Hladic)

• Criminals May Adopt Faster
  Less skilled, resource-constrained criminals may tap AI agents as force multipliers (e.g., for basic automation in ransomware campaigns).

• No Silver Bullet: Prevalence Will Rise But Not Overnight
  Joe predicts eventual, broader criminal use once barriers fall—but immediate widespread autonomous attacks are unlikely.

  "Agentic AI is a technique that I think is going to be used... but until it becomes viable for threat actors... that's an inevitability. I just don't know when that timeline is." (18:32, Joe Hladic)

Preparation and Vigilance: The Defender’s Perspective

• Defenders Must Get Ahead
  While timing remains uncertain, security teams can't ignore the risks posed by AI-driven identities or agents.

  "We should put a lot of our energy and focus into [AI agents], to understand how they can be exploited, how they're vulnerable, how they can be tampered with and manipulated." (21:38, Joe Hladic) "That's a big command and control space to deal with, and we have to understand it in order to defend it." (21:57, Joe Hladic)

Notable Quotes & Memorable Moments

• "If you own the backups, you own the business. Because if you own the backup data, there's no way they can recover."
  — Joe Hladic (12:17)

• "The perimeter is no longer the network, identity is the new perimeter."
  — Joe Hladic (03:36)

• "For every human identity, there are 82 non-human identities."
  — Joe Hladic (04:14)

• "Agentic AI is a technique that I think is going to be used, but until it becomes viable for threat actors... that's an inevitability. I just don't know when that timeline is."
  — Joe Hladic (18:32)

• "AI still hallucinate[s]. Humans, in my opinion, are still a lot more trustworthy... that's too much of a risk for a lot of threat actors to actually do."
  — Joe Hladic (15:37)

Timestamps for Critical Segments

| Segment | Time | |--------------------------------------------|------------| | Identity: The New Perimeter | 03:11–04:40| | Non-human Identities & New Attack Surface | 04:10–05:48| | Prioritizing Recovery (Order of Ops) | 06:15–09:08| | Stats: Shrinking Recovery Confidence | 09:08–09:54| | Why Enterprises Still Pay Ransom | 09:54–10:54| | Backup Systems: Threat Actor's Leverage | 12:17–13:51| | Agentic AI Adoption: Myths & Reality | 13:51–17:43| | Will Attackers Use Agentic AI Soon? | 19:05–21:24| | Defender's Mandate: Vigilance on AI Agents | 21:34–22:03|

Conclusion

This episode delivers a grounded, practical, and occasionally sobering look at the challenges facing modern cybersecurity teams. The identity landscape is evolving rapidly, with a dramatic rise in non-human identities and early, sometimes overhyped, adoption of agentic AI. Organizations struggle not just with prevention, but with prioritizing effective recovery—especially as attackers shift to targeting backups directly. Joe Hladic's core message: resilience strategies must be business-specific and proactively inclusive of future AI-driven threats, even if the hype sometimes outpaces the palpable risk.

Find the full Rubrik Zero Labs “Identity Crisis” report and more insights: zerolabs.rubrik.com