The Watchtower

Are we building enterprise AI backwards? As organizations race to adopt bigger, more powerful AI models, Alex Attumalil, Global CISO at Under Armour, believes the real transformation may come from thinking smaller.
In this episode of The Watchtower, Alex joins Ash Hunt to explore why general-purpose LLMs aren't enough to transform complex business processes, how smaller, specialized AI models could reshape the enterprise, and why business leaders, rather than technology teams, should be driving AI adoption.
They also examine how AI is changing cybersecurity itself. From sophisticated phishing attacks that are forcing organizations to rethink decades of security training to AI-powered security operations that can deliver faster, more consistent decisions, Alex shares how security leaders must adapt to a rapidly changing threat landscape.
And as automation continues to evolve, what happens when AI can identify software vulnerabilities before they ever reach production?

Connect
Alex Attumalil on LinkedIn
Under Armour
Ash Hunt on LinkedIn
The Watchtower

Learn More
datasecAI by Cyera

Chapters
0:00 One giant brain, or lots of small ones?
1:18 Meet Alex Attumalil, Global CISO at Under Armour
2:19 The two-front war: defending AI, and defending against AI
4:47 The decade of phishing training that just became obsolete
7:32 Running security consistently across a global business
10:38 The only SecOps metric that matters: time to action
15:32 'Democratizing the response' — AI's hidden payoff
16:53 Why your tech team is the wrong owner for AI
22:44 The micro-LLM thesis: enterprise AI as microservices
24:53 Repatriation: bringing tuned AI back into the data center
25:30 Why LLMs keep giving you different answers
28:14 SLMs are quietly beating LLMs in the enterprise
30:52 Where the CISO role goes next
34:44 Could software ever have no vulnerabilities?
37:05 Every enterprise does two things. The rest is scaffolding.

Presented by Cyera. Produced by Mission.org.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

If Convenience Wins, We Lose.
Evan Gordenker led more than 160 insider investigations at Unit 42, including work on DPRK IT-worker operations, before taking over insider threat and data security at Airwallex.
Evan joins Ash Hunt to showcase how intent is often the wrong place to start. Whether the risk comes from an employee, an AI agent, or an adversary using legitimate credentials, the real question is what trusted access allows them to do.
Evan explains why convenience sits at the center of human and agentic risk, how AI agents expose the limits of access models built for people, why security teams need finer-grained control over credentials and permissions, and why the DPRK IT-worker playbook keeps working without changing much at all.

Connect
Evan Gordenker on LinkedIn
Airwallex

Chapters
0:00 The insider threat you're worried about isn't a person
1:19 How Evan fell into cybersecurity via the Tokyo Olympics
4:20 What the industry gets wrong about who the insider actually is
5:25 Why the 'secret police' framing kills your program
8:00 Intent doesn't matter — controls do
9:30 What a real, messy insider investigation actually looks like
12:30 One sentence in an incident report can end a career
15:01 Would you classify an AI agent as an insider?
16:33 The one-password experiment — how agents test every credential
19:39 4 years into agentic systems. Still built for humans only.
23:54 If security runs insider risk alone, you're toast
28:35 The DPRK IT-worker threat — and why it hasn't had to evolve
32:54 How Fortune 500s started frustrating North Korean operators
36:48 The one through-line: convenience
42:50 Why InfoSec has never mattered more than right now

Presented by Cyera. Produced by Mission.org.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Should detection and response remain the center of security when AI, data, and identity are converging at machine speed?
Detection and response has long been an organizing principle of enterprise security. Francis Odum, Founder and CEO of Software Analyst Cyber Research, argues that model is being pushed to its limits as AI agents act autonomously and attack timelines compress.
In this episode, Francis and Ash Hunt explore why existing identity controls may not be enough for AI agents, why every AI security vendor sounds identical right now, and how security controls have to change for agentic systems.


Connect
Francis Odum on LinkedIn
Software Analyst Cyber Research
Ash Hunt on LinkedIn


Chapters
0:00 Can Human-Speed Security Stop Machine-Speed Threats?
1:15 Francis Odum on What's Changing in Cybersecurity
3:53 Cybersecurity Was Built for Humans — Not Agents
8:44 Why Every AI Security Vendor Sounds the Same
11:08 Why Platforms May Win the AI Security Race
14:31 What Happened to the First Wave of AI Security Startups
19:20 Why Non-Human Identities Were a "Head Fake"
22:00 Should AI Agents Have Their Own Identity?
25:58 The Authorization Problem AI Agents Create
28:29 What Hugging Face Revealed About Agent Controls
32:15 Can Detection and Response Keep Up With AI?
34:27 How AI Makes Attacks Faster and Cheaper
36:02 Why Security Has to Become More Preemptive
41:03 The New Agentic Supply Chain Risk
43:27 Why Agentic Failures May Not Look Like Cyberattacks
45:13 What SACR Is Researching Next



Presented by Cyera. Produced by Mission.org.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

If you focus entirely on the external AI-fraud threat, you're going to get burned by what's already inside your organization.

AI is making external fraud more convincing - deepfake CEOs on service-desk calls, synthetic video, generative BEC. But the same tools also scale internal fraud, from fake invoices and fabricated expenses to ghost employees.

Mason Wilder, Research Director at the Association of Certified Fraud Examiners, joins Ash Hunt to unpack why it's never been easier to commit fraud, how organized crime is changing the landscape, and why the old walls between cybersecurity, fraud, and AML are starting to break down - and where companies should actually be investing to keep up.

Connect

Mason Wilder on LinkedIn

ACFE

ACFE Anti-Fraud Technology Benchmarking Report

ACFE Report to the Nations

Chapters

0:00 The AI Fraud Threat Inside Your Company

1:03 From Overnight SOC Shifts to the ACFE

2:56 Why Fraud Has Never Been Easier

4:55 What the 2026 AI-Fraud Survey Revealed

5:25 Why Only 7% Feel Prepared for AI-Powered Fraud

6:12 Why Cyber, Fraud, and AML Can't Stay Siloed

9:33 How Deepfake Fraud Started in 2018

11:54 From Social Engineering to Automated Attacks

13:41 Where Companies Should Invest Against AI Fraud

16:53 The Internal Fraud Risk Getting Overlooked

20:00 Why Internal Fraud Doesn't Go Away

22:52 Why Tips Still Catch 43% of Fraud

26:18 Turning Employees Into a Fraud Detection Layer

28:17 What Agent-to-Agent Fraud Changes

34:59 How Singapore Is Shifting Fraud Liability

37:11 What Deepfake Detection Really Looks Like

42:24 Why Cyber and Fraud Are Ultimately on the Same Team

Presented by Cyera. Produced by Mission.org.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

When security deals in absolutes, the business finds a way around it.

Bose CISO Arun Abraham joins Ash Hunt to discuss why flat “no”s create shadow IT, how security leaders can enable AI without opening the gates, and why fundamentals like data, identity, vulnerability management, and risk prioritization still matter most.

 

Connect

Arun Abraham on LinkedIn

Bose Corporation

The Watchtower

 

Chapters

0:00 Are We Too Secure?

1:25 The CISO outsider advantage

5:19 Puritanism in security

6:24 How to earn trust when you tell executives the truth

8:11 Why quantitative risk beats maturity curves

9:39 The three fundamentals: detect, respond, VM

13:44 What AI enablement requires from security

14:14 The cloud playbook applies: enable or get shadow AI

15:10 The CISO who got replaced for saying no

15:49 Why Bose's security team has a backlog

19:59 The five-minute aha moment

21:34 If you don't know where your data is, you can't enable AI

25:09 The internal AI agent that replaced policy

27:22 The secure SDLC didn't look like this five years ago

29:44 AI won't invent new attacks - it'll use your deficiencies

31:04 Perfect is the enemy of good

Presented by Cyera. Produced by Mission.org.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.