An employee typed his own name into Copilot and found out he was on the layoff list.

Nobody hacked anything. Years earlier someone had used "share with all" in SharePoint, back when nobody could realistically guess the link. Then AI made everything trivially findable.

Ward Balcerzak is Field CISO at Sentra with nearly two decades in data security across Accenture, Allstate, Carbon Black, and Fidelity National Financial, where he led data protection and insider risk programs. He also hosts the Guardians of the Data podcast.

Aaron and Ward get into what data security looks like now that every employee is pasting company data into tools nobody approved. Ward's argument: the work organizations skipped for years — data discovery, classification, access hygiene — is exactly what AI is now dragging into the light.

He also makes a contrarian case for protecting less. Nine times out of ten, the data companies are frantic about isn't special at all. Figure out what actually makes you different, find out who really has access to it, and start there.

They cover why DLP earned its bad reputation, what to do when your tools hand you a million findings and no way to act on them, why regulators always arrive last, and why networking is now non-negotiable even for people who got into this field specifically to avoid people.

Guest: Ward Balcerzak, Field CISO at Sentra. Find him on LinkedIn, at wardbalcerzak.com, or on the Guardians of the Data podcast.

⏱️ CHAPTERS
00:00 Intro
01:25 Twenty years of being wrong about what catches on
04:39 Can AI actually save data security?
09:30 The era of YOLO security
12:00 "AI readiness" and why nobody's ready
14:43 Sins of the past, now with petabytes
17:06 Are companies building their own models?
26:30 It's not a cat anymore, it's a mountain lion
28:16 The rigged casino of AI watermarking
29:24 Where are the actual wins?
30:59 Nine times out of ten, your data isn't special
33:48 Who actually has access?
35:09 The Copilot layoff list story
37:06 A million findings and no idea what to do
45:33 Forcing the conversations nobody wants to have
47:21 What to tell a 22-year-old today
50:26 Network, even if you got into this to avoid people
54:47 Pitch slapping and the value of being real
56:03 Where to find Ward

#cybersecurity #datasecurity #AI #infosec #CISO

Every CISO or engineer who's ever been told no by a board should watch this.

Joel Benge spent years as the communications lead for the Department of Homeland Security's CISO office before writing "Be a Nerd That Talks Good." He's spent his career on one problem: getting technical people the budget, the headcount, and the buy-in they deserve — and figuring out why they so often don't get it.

Aaron and Joel get into why most board decks fail before slide three, the "blank stare moment" that happens when you overload your audience, and the reframe that changes everything: you're not the hero of your pitch — you're Obi-Wan, and the person you're asking is Luke. Plus why the stories you least want to tell (the failures, the things you tried that didn't work) are the ones that actually build trust.

They also get into why every vendor booth at Black Hat says the exact same thing, why "we prevent hacks" is a Gartner category and not a big idea, and how to align a security ask to something the business already cares about — instead of another slide full of blocked-attack metrics nobody reads.

If you've ever needed a million dollars for a program and walked out with nothing, this is the episode.

Guest: Joel Benge, author of "Be a Nerd That Talks Good." Find him at nerdthattalksgood.com or on LinkedIn — the intro and first chapter are free, no email required.

⏱️ CHAPTERS
00:00 Intro
02:09 Where the community is right now
06:13 Standing out in the sea of sameness
12:00 Stop copying each other's homework
13:10 "I can tell your content is AI"
15:08 "Live, laugh, love" and the Gartner-category trap
17:30 Mind, gut, heart
25:53 The stories founders refuse to tell
30:12 Why heart is what cyber gets worst
32:22 Finding the real big idea
34:09 Second- and third-order benefits
38:11 CISOs, boards, and getting budget
39:04 The blank stare moment
48:24 You're Obi-Wan, not the hero
52:35 We made CEOs the heroes and forgot customers
55:00 Aligning security to what the business values
56:06 "More scared of their company than the bad guys"
57:18 Where to find Joel

#cybersecurity #CISO #infosec #leadership

"You know who will be hiring juniors again? Bad guys."

Casey Ellis founded Bugcrowd and launched the first bug bounty programs on it back in 2012, pioneering crowdsourced security as a service. He co-founded disclose.io, sits on the Black Hat and DEF CON Policy review boards, and now runs Tall Poppy Group, angel investing and advising the next generation of security startups.

Fresh off a week at Black Hat, B-Sides, and DEF CON, Casey joins Aaron for a wide-ranging conversation about what he actually heard on the ground — and why the mood at each of those three conferences was completely different.

We get into the "slopdemic" (Casey's term for AI-generated vulnerability reports drowning the ecosystem), why he's "human pilled" rather than AGI pilled, and his read on the White House memorandum he's calling the privateering order — what it actually authorizes, and who's really going to use it. Plus: why pen test firms are about to have a very hard time defending their value, why every company already has a vulnerability disclosure program whether they know it or not, and the hygiene fundamentals that still contain the blast radius when everything else fails.

The last stretch is the one worth staying for — a genuinely urgent case for why the industry's "we're never hiring juniors again" moment is a gift to the people recruiting them instead.

Guest: Casey Ellis, founder of Bugcrowd and disclose.io, principal of Tall Poppy Group. Find him on LinkedIn.

⏱️ CHAPTERS

00:00 Intro

02:35 Coming out of Black Hat, B-Sides, and DEF CON

06:33 Why DEF CON was allergic to the AI hype

09:35 Hybrid conflict is already here

10:53 Royalty in the palace, villagers at the gate

12:16 Security below the poverty line

13:14 "I'm not AGI pilled. I'm human pilled."

14:51 Do stupid things faster with more energy

19:04 What people get wrong about AI and exploitation

21:43 The slopdemic and the vulnpocalypse

25:07 What it takes before anything actually changes

28:48 Nobody is coming to save you

33:02 What actually works if you start from scratch today

37:34 Why pen test is in for a ride

39:13 Hygiene, blast radius, and the boring basics

43:31 The privateering memorandum

48:26 Who's actually waiting to hack back?

51:24 What to tell a 22-year-old today

53:21 It's really easy to do crime

55:08 Community, disclose.io, and knowledge transfer

58:01 The industry needs to give back

60:45 "Who will be hiring juniors again? Bad guys."

61:11 Tall Poppy Group and where to find Casey

#cybersecurity #infosec #AI #bugbounty #DEFCON

I've been going to BlackHat since 2001. This year, the scariest thing on the floor was a booth almost nobody was crowding.

Fresh off plane from Vegas, this is a solo ZeroSum episode — no guest, just my top 5 takes from the week. The throughline: the industry is running on hype and extraction, while the one real capability shift sat in plain sight.

We get into build vs. buy flipping, the "ask me about Mythos" cargo cult, how transactional the show floor has become, the total absence of anything new, and the modest OpenAI booth that quietly made the case it could eat half the vendors in the hall.

Chapters:
00:00 Intro — 25 years of Black Hat, from 2001
02:24 Take 1: "We'll just build it ourselves"
04:48 Take 2: Ask me about Mythos
06:52 Take 3: Everyone's selling the same thing
17:00 Take 4: 400 booths, zero controversial statements
18:58 Take 5: The Amazon-in-the-corner booth

Listen anywhere — search "ZeroSum" on Spotify and Apple.

What did you take away from Black Hat this year? Drop it in the comments.

#BlackHat #Cybersecurity #AI #InfoSec #ZeroSum

Malware used to take 18 minutes to move through your network. Now it takes 27 seconds. Humans can't respond fast enough anymore.

Graham Westbrook is VP of International Markets at SimSpace — the "AI Proving Grounds" for cybersecurity — with a background spanning DoD/DISA, CrowdStrike, Flashpoint, and a postgrad in AI from Oxford.

Aaron sits down with him for a genuinely global, big-picture conversation about where AI security is actually heading — not from Silicon Valley, but from the ministries of defense, critical infrastructure operators, and enterprises he works with around the world.

We get into why "legacy SOC is no longer sufficient" and the shift from human-in-the-loop to AI-first defense, why governments and banks will build their own models on their own data instead of renting from OpenAI or Anthropic, and the case for preemptive defense — forecasting and patching attacks before they happen, the way Google forecasts weather. Plus: the "zero day clock" racing toward minutes, why finding more vulnerabilities is a dead end, whether AI's benefits will ever reach smaller organizations or just the giants, and why Graham thinks an AI market crash is coming — and what that means for anyone entering the field.

Guest: Graham Westbrook, VP International Markets at SimSpace. Find him on LinkedIn @ https://www.linkedin.com/in/graham-westbrook/

⏱️ CHAPTERS
00:00 Intro
01:57 Where the world actually is on AI (hint: behind)
04:06 Battlefield AI and preemptive cyber defense
06:23 Why every nation will need its own model
09:48 AI, layoffs, and the "data shepherd" future
11:12 What enterprises are actually doing today
14:14 The 27-second breakout and why humans can't keep up
16:09 The zero day clock and forecasting attacks like weather
24:58 Building a digital replica of your organization
29:33 Will AI's benefits ever reach the little guy?
32:49 The arms race with China
55:37 Getting high on your own supply — the coming crash
56:42 Advice for anyone entering cybersecurity

He's an L8 Principal Engineer at Netflix — the kind of technical leader the company puts on the problems that decide whether it wins or loses. And he just watched AI out-hack him.

Scott Behrens is an L8 Principal Security Engineer at Netflix, where over 11 years he's watched the security team grow from a handful of people to over a hundred. Today he's the technical lead for Netflix's Live product security, its Attack Emulation Red Team, and its DDoS research. He joins Aaron for one of the most honest, forward-looking conversations we've had about what AI actually does to security work, and to the people who do it.

Scott doesn't sugarcoat it: he sat down with the latest models and quickly concluded they're better at finding and exploiting vulnerabilities than he is. But instead of doom, he lays out why that's an opportunity and where humans still hold the irreplaceable edge.

We get into why the model matters less than the "harness" you build around it, the idea that human intention and hard-won wisdom are the most valuable resources in the AI race, and what actually happens when your discovery tools start surfacing thousands of real vulnerabilities you now have to fix. Plus: how AI is quietly making security the easiest story he's ever had to tell, the one-line trick that cuts vulnerabilities in AI-written code, and why the best security engineers are becoming systems thinkers, not bug-finders.

Guest: Scott Behrens, L8 Principal Security Engineer at Netflix.

Find him on LinkedIn and read his newsletter, The Engineer Setlist, on Substack.

⏱️ CHAPTERS
00:00 Intro
01:57 Excited, worried, and humbled all at once
04:16 Don't just do the old things faster
07:30 Should security teams fix the bugs, not just find them?
09:30 Human intention is the most valuable resource in the AI race
10:16 The "wisdom" AI doesn't have
12:22 Systems thinking as the human edge
18:14 Why the harness matters more than the model
20:25 Codifying 20 years of expertise into a harness
23:41 You built the harness — now what do you do with the findings?
25:32 What small and mid-size businesses should actually do
27:35 The one-line trick that cuts vulnerabilities in AI code
30:41 Rethinking the front end, WAFs, and detection
32:45 The "isadmin=false" honeypot trick
51:50 The era of YOLO security
53:09 Security as an enablement function
55:04 "The easiest story I've ever had to tell"
56:34 Where to find Scott

#cybersecurity #infosec #AI #Netflix #appsec

For 30 years, the security industry could never fully win. Justin Somaini explains what finally changes that.

Justin Somaini is one of the people who helped define the modern CISO role — former CISO of Symantec, Yahoo, SAP, and Box, and now a Partner at YL Ventures. In this episode, Aaron sits down with him for a genuinely optimistic conversation about why security has always fallen short, and why he believes we're at the most exciting inflection point in the industry's history.

Justin lays out his core thesis: security has never truly succeeded because of two structural problems — you can never hire enough people to review everything, and the industry never solved the "shared accountability" gap with engineering and IT. For the first time, he argues, AI can absorb the enormous amount of pattern-matching work that's always overwhelmed security teams — letting them finally operate at the scale the business actually demands.

We also get into how he learned to separate real founders from the hype (and the belief he had to unlearn), why "drop the first slide" is his number one pitch advice, how CISOs actually make buying decisions versus what they claim, and his honest take on where the funding market is heading. Plus: the first CISO ever, why word-of-mouth beats every marketing tactic, and what makes a startup worth betting on.

Guest: Justin Somaini, Partner at YL Ventures. Find him on LinkedIn or on his podcast, Somaini Trust Issues.

⏱️ CHAPTERS
00:00 Intro
01:38 Learning from Steve Katz, the first CISO ever
04:03 Why a 30-year veteran is optimistic right now
05:23 Why security has always failed — the two real reasons
07:42 How AI absorbs the work that overwhelms security teams
10:03 How to separate real founders from the hype
13:13 Salesperson, or a product engineer faking it?
16:15 "Drop the first slide" — fixing the founder pitch
18:20 Why first-time founders are like teenagers
22:10 The funding market and where it's headed
28:12 Need-to-have vs. nice-to-have
36:42 How CISOs actually make buying decisions
52:00 How to get anyone to care about what you're building
57:04 Where to find Justin

The haystack got bigger. The needles didn't. Ed Bellis on why the vulnerability problem stopped being human-scale.

Ed built Kenna Security, sold it to Cisco, and is now back with Empirical Security, building custom AI models that predict which vulnerabilities will actually get exploited in your environment — not just which ones score high on a generic global list. He joins Aaron for a candid conversation about what AI is really doing to the vulnerability landscape, the security talent pipeline, and the funding environment.

We cover why AI is producing more code than ever and more vulnerable code than ever, what Anthropic told him about not hiring junior devs anymore ("what happens when the seniors time out?"), why every organization needs a model trained on its own environment, and where startups still win against Microsoft and Google. Plus: why raising $180M in a seed round quietly kills your optionality, the "fast no," and the Chicago startup problem.

Guest: Ed Bellis, CEO of Empirical Security → empiricalsecurity.com

⏱️ CHAPTERS
00:00 Intro
01:20 Coming off a raise — what it actually changes
03:22 What money doesn't fix
05:02 The mistakes he won't repeat from Kenna
07:47 Why a huge seed round kills your optionality
09:03 Is private equity coming for cyber?
12:48 Why every company needs its own model
17:22 Is it just YOLO security now?
18:28 The haystack got bigger, the needles didn't
20:25 AI, remediation, and operational risk
23:45 Anthropic doesn't hire junior devs anymore
26:19 If you take the bottom rung out of a ladder
30:42 Advice for a 22-year-old entering cyber
32:57 Cutting through the noise at Black Hat
36:47 Why he loves a "fast no"
37:26 The Chicago startup problem
39:22 Where startups fit against Microsoft and Google
43:47 Build vs. buy: should you train your own models?
49:02 Where to find Ed

#cybersecurity #infosec #AI #vulnerabilitymanagement #startups

"We don't need to throw out what we've been doing all of our careers. We just need to be AI-aware in how we apply it."

Diana Kelley — CISO at Noma Security, formerly CISO at Protect AI, and a 25+ year veteran of Microsoft, IBM, and Symantec — joins Aaron for a grounded conversation on what actually changes, and what doesn't, when AI enters security.

We cover why trust boundaries and identity are still the job even as AI reshapes the attack surface, why the LLM itself can't function as a trust boundary, the shadow AI sprawl most CISOs haven't gotten a handle on, and the identity mess created by agents inheriting permissions from whoever spins them up. Plus: does AI actually level the playing field for builders, or just make the biggest players bigger?

Guest: Diana Kelley, CISO at Noma Security.

⏱️ CHAPTERS

00:00 Intro

01:34 Nearly four decades in security — where are we right now

13:23 The labs stole the data to begin with

22:19 Shadow AI sprawl and the identity problem

28:21 First principles: trust boundaries, not a full rewrite

39:18 Are the frontier labs going to crush every startup?

42:52 AI, equality, and who gets left behind

55:56 Where to find Diana and get involved

#cybersecurity #CISO #AI #infosec

Is any security company safe from AI disruption? Evolve Security CEO Mark Carney's answer: no and that's the wrong question anyway.

Aaron talks with Mark Carney, CEO of Evolve Security and one of the OG figures in security services, about where the industry sits at a genuine inflection point. Mark breaks down why tech-enabled services are hotter than ever while human-only services are a dying breed, and how private equity now runs "AI disruption due diligence" before writing big checks.

We get into:

• Why finding vulnerabilities became a commodity — and where the real value lives now
• Why Evolve deliberately isn't using AI for autonomous exploitation yet
• The 50% false-positive problem with autonomous pen testing tools
• Attack chains, business logic, and prioritizing high-impact exposures
• Chartering an "autonomous enterprise" across every business function
• The death of tasteful outbound and the abuse of the CISO relationship
• Career advice for a tough market: deliver outcomes, not activity

Guest: Mark Carney, CEO of Evolve Security. Find him at evolvesecurity.com or on LinkedIn.