Talkin' Bout [Infosec] News

Recorded live at Wild West Hackin’ Fest, the BHIS crew debates AI agent sandbox escapes, OpenAI’s frontier model decisions, and accountability for the Hugging Face hack. Other stories include the dismissal of OpenAI safety researchers, the FBI–Accenture breach, ShinyHunters arrests, Oracle’s medical records breach, and recurring Citrix vulnerabilities. The panel explores corporate scapegoating, stolen authentication tokens, patching challenges, and Anthropic’s expanding cybersecurity access—asking who takes responsibility when security controls fail.

Join us LIVE on Mondays, 4:30pm EST.
A weekly Podcast with BHIS and Friends. We discuss notable Infosec, and infosec-adjacent news stories gathered by our community news team.
https://www.youtube.com/@BlackHillsInformationSecurity

Chat with us on Discord! -
https://discord.gg/bhis
🔴live-chat


Chapters
  • (00:00) - Live From WWHF Deadwood '26 - 2026-10-07
  • (01:13) - OpenAI’s Frontier Model Pause and AI Sandbox Escapes
  • (05:42) - OpenAI Dismisses Three Safety Researchers
  • (07:20) - FBI–Accenture Breach and Blaming Individual Employees
  • (09:33) - Oracle’s Medical Records Breach and Cloud Responsibility
  • (11:12) - ShinyHunters Arrests and the FBI Breach
  • (16:54) - Critical Citrix Vulnerabilities and Incomplete Patches
  • (19:28) - Selling Stolen VPN Access and Authentication Tokens
  • (22:49) - Anthropic Expands Its Cyber Verification Program
  • (27:23) - Lawsuit Over the Hugging Face Hack and AI Accountability

Links
OpenAI’s Frontier Model Pause and AI Sandbox Escapes
OpenAI Dismisses Three Safety Researchers
FBI–Accenture Breach and Blaming Individual Employees
Oracle’s Medical Records Breach and Cloud Responsibility
ShinyHunters Arrests and the FBI Breach
Critical Citrix Vulnerabilities and Incomplete Patches
Anthropic Expands Its Cyber Verification Program
Lawsuit Over the Hugging Face Hack and AI Accountability


Click here to watch this episode on YouTube.




🔗 Register for FREE Infosec Webcasts, Anti-casts & Summits 
https://poweredbybhis.com

Brought to you by:
Black Hills Information Security 
https://www.blackhillsinfosec.com

☯️ Introducing BHIS Fusion Penetration Testing
https://www.blackhillsinfosec.com/fusion-penetration-testing/

Antisyphon Training
https://www.antisyphontraining.com/

Active Countermeasures
https://www.activecountermeasures.com

Wild West Hackin Fest
https://wildwesthackinfest.com

Creators and Guests

Host
Bronwen Aker
Bronwen Aker is a BHIS Technical Editor who joined full-time in 2022 after years of contract work, bringing decades of web development and technical training experience to her roles in editing pentest reports, enhancing QA/QC processes, and improving public websites, and who enjoys sci-fi/fantasy, Animal Crossing, and dogs outside of work.
Host
Corey Ham
Corey Ham has been with Black Hills Information Security (BHIS) since 2021 delivering red teaming and OSINT services. Currently, Corey leads the ANTISOC team at BHIS, providing subscription-based continuous red teaming to BHIS clients. Outside of his time at BHIS, you can find him out in the woods or up on a mountain somewhere.
Host
John Strand
John Strand has both consulted and taught hundreds of organizations in the areas of security, regulatory compliance, and penetration testing. He is a coveted speaker and much loved SANS teacher. John is a contributor to the industry-shaping Penetration Testing Execution Standard and 20 Critical Controls frameworks.
Host
Ralph May
Ralph is a U.S. Army veteran and former DoD contractor who supported the United States Special Operations Command (USSOCOM) with information security challenges and threat actor simulations. Over the past decade, he has provided offensive security services at Optiv Security and Black Hills Information Security (BHIS) across various industries. His expertise spans network, physical, and wireless penetration testing, social engineering, and advanced adversarial emulation through red and purple team assessments. Ralph has developed several tools, including Bitor (set to release in January 2025) and Warhorse, which enhance efficiency in penetration testing infrastructure and operations. He has spoken at numerous conferences, including DEF CON, Black Hat, Hack Miami, B-Sides Tampa, and Hack Space Con.
Guest
Andy Pettit "Nerf"
Andy Pettit is a cybersecurity practitioner and lifelong builder with a hacker’s mindset, driven by deep curiosity and a desire to understand how systems truly work. He began coding in C at age 12 building custom MUDs and has been pulling systems apart ever since, focusing on gaps between design and real-world behavior. Andy brings a whole-business perspective from over a decade as managing partner of Clown Shoe Motorsports, shaping his views on risk, reliability, cost, and people. He volunteers with Black Hills Information Security and Antisyphon Training as a Nerd Herder and is a top 5% MetaCTF competitor, endurance racer, and HPDE instructor with NASA Texas Region.
Guest
Jake Williams
Adrien Lasalle, known online as NetRunner, is the founder of NetRunSecurity and an offensive security professional with nearly seven years of experience breaking things professionally, from web applications and Active Directory to cloud, OT/SCADA, and embedded devices. His specialty is hardware hacking, PCB design, and anything with a chip in it. A keynote speaker at NorthSec 2025, Adrien has also presented at Wild West Hackin' Fest Deadwood, InCyber Canada, and BSides events across Montreal, Warsaw, Kraków, and Copenhagen. A former firefighter in France, he is now based in Montreal, where he shares his passion for hardware, cybersecurity, and pentest war stories and memes weekly on LinkedIn and Twitch.

What is Talkin' Bout [Infosec] News?

A weekly Podcast with BHIS and Friends. We discuss notable Infosec, and infosec-adjacent news stories gathered by our community news team. Join us live on YouTube, Monday's at 4:30PM ET.

Corey Ham:

What is...

John Strand:

I need the finger. Finger? Roll out the finger. Feel a little left out here.

Corey Ham:

John, you should come come over at least to this. Yeah. I I feel like it's us versus John.

John Strand:

It's okay with you, I guess.

Andy Pettit:

I I think us versus John is a great way to do the news.

John Strand:

I got you. You you

Ralph May:

don't have one? There's like six over there.

Bronwen Aker:

Thank you, sir. You, sir. You're a gentleman and a scholar and I promise not to tell a soul. Oops.

John Strand:

So good at lying. Alright.

Corey Ham:

Alright. Someone give us a thumbs up if we can do the finger. I I can't...

John Strand:

Do the intro. We got that. We got a thumb up. We got the finger coming.

Corey Ham:

Do the finger.

John Strand:

Oh my god. It's us. Why did anyone tell me I was that fat?

Ralph May:

Just got a

John Strand:

look. I hate seeing myself in camera.

Corey Ham:

Okay. So there's basically two articles. Two. And there's actually like a thousand. But the OpenAI...

Ralph May:

We can only count that high too.

Corey Ham:

Corey, we can't count that high. You guys know this if you've ever watched the show. So there... The first article is OpenAI announcing that they're not going to release their latest frontier model. And also that they're not gonna continue training frontier models.

John Strand:

Because it's it's too perilous.

Andy Pettit:

I think we can handle the peril.

John Strand:

That... You wanna face the peril?

Andy Pettit:

Just a little bit of peril. Just a little

John Strand:

bit of peril. You think you can handle it? And then the spankings.

Corey Ham:

So the AI has been doing a little too many spankings. It's basically the problem.

John Strand:

Because I

Ralph May:

I feel like OpenAI was just like, you know, we're getting our ass kicked, so we're just gonna slow down.

Corey Ham:

We're gonna give up? Okay. So there are a few, like, credible things that happened. They had, like, a attempted DNS x fill thing they posted about that was...

Ralph May:

Every sandbox seems to be broken with AI. Correct. Everyone.

Corey Ham:

Yeah.

John Strand:

Okay.

Bronwen Aker:

Okay. In fairness, last night while I was trying to get to sleep, I logged on to Perplexity and I said, look, we've had all of these agents going through doing things they shouldn't do. I want a full detailed analysis. I got 11 pages. And the commonalities in almost every one of the escaped AI agents was operator error.

Bronwen Aker:

They had removed safeguards. They had broken their own stuff. And then they said, go get them.

Corey Ham:

So this one was a DNS. Basically, the agent attempted to tunnel over DNS. That's classic. Classic. I mean, it's such a good pen tester.

Corey Ham:

It's such an, you know, DNS tunneler. But it got caught. They... It's pretty funny how they're doing a little bit of virtual... Virtue signaling with the whole, like, our sandboxes work now.

Corey Ham:

Because in the blog, they say, okay. It got caught within three minutes, and an operator reviewed it within fifteen minutes. Then they also say, and the run was killed two and a half hours later. It's like, guys, that's not... You're you're you're not really flexing your SLAs on us.

Corey Ham:

It's kinda the opposite. You're saying like, yeah. I mean, we let it do crime for like a couple hours, but not like two months. Okay. It's the...

John Strand:

Oh, we got Jake over here.

Jake Williams:

Hi, Jake. Can I counterpoint on this? Yes. Yes. So I guess that's actually what this whole thing is about.

Jake Williams:

It's your whole... Literally the whole jam. Right? So the the counterpoint on this is that those eval's are hella expensive. Right?

Jake Williams:

And so, you know, that two and a half hour, I believe it was. It was two and a half hours, Seth. Right? I know it was a couple hours. Right?

Jake Williams:

So that couple hours, you know, space there between the detection and them deciding to terminate the, you know, certainly terminate the the run. Right? I think that somebody doing evaluations around, you know, how bad is this? Has this truly poisoned the well, as it were, right, to the point that this is no longer a valid test? There's a real cost thing going on here.

Jake Williams:

And I'd like to believe, although maybe I shouldn't, right? But I'd like to believe they were monitoring, right, during those two and

Corey Ham:

a half hours.

Ralph May:

Is it an AI monitoring though?

John Strand:

So I agree with you right up into that last point. And I'm gonna rip on Silicon Valley here for a little bit. Talked about it when we were teaching in... Aker. Where were we?

John Strand:

Was it San Francisco or San Jose? Anyway, so if you are in in the universe of computer security in The United States, you have The United States and New York is very different, And then Silicon Valley is incredibly different. And I'm just gonna be blunt. A lot of the security professionals in Silicon Valley have their heads so far up their ass, they will never see daylight again. Right?

John Strand:

They have the cool facial hair. They have the awesome skinny jeans. And they have an attitude that they are a 10 x engineer. And there's so much hubris in that town that whenever you're talking about anything, or whenever we were training them, they would be blatantly wrong about something. And they would just go down in flames all the way.

John Strand:

It's just... It's an attitude in that area. And whenever I hear about this escaped, and it used DNS tunneling, and we're like, why wouldn't anybody be looking for that? Because of the people that are running the security are idiots. I could try to try to, like, dance around it.

John Strand:

I'm just gonna say, we've been there. We've talked to them. We've argued with them for hours. They're absolute morons. And because they have so much...

John Strand:

So many shares, and they're worth 10 times more than we'll ever be worth, they feel like they don't have to listen to, like, basic security principles.

Corey Ham:

So there are three people who got fired from OpenAI. Do we think they're, like, security experts? They're the ones...

John Strand:

I bet you they lined them up based on how cool they looked. They got rid of the ones that were like, these guys don't look like they're from here.

Jake Williams:

This is an article that got created. What's Where's your Patagonia vest?

John Strand:

Yeah. Where's your Patagonia vest? Exactly.

Corey Ham:

This is an article in the Hacker News. Basically, they parted ways with three safety researchers Oh. Over sensitive information mishandling is what they're saying. I don't know. There's not clear details on whether they're like whistleblowers who said like, hey, we're like hacking Hugging Face.

Corey Ham:

Like, is this okay? Like, I don't... We don't know the details on whether they're whistleblowers, whether they're just people leaking things to the press. But, I mean, read between the lines here. Right?

Corey Ham:

It's like, they have all these safety issues and now there's three people who have been dismissed. Maybe they're like, hey, stop leaking.

John Strand:

Wanna with... I wanna go back to that thing. They were dismissed from OpenAI. I thought they had a third party firm that was doing the security for this.

Andy Pettit:

Well, for the testing, yeah. They were having... Meter.

Jake Williams:

With the external firm that's looking at Meter.

Andy Pettit:

Yeah. I'm blanking on the name. Right? But it's been the same firm for like all of them. I was gonna say on the idiot level that John was talking about, if these guys actually x filled sensitive data themselves instead of getting an agent to do it that they

Corey Ham:

could just blame. Yeah. They deserve to get fired. Was I don't know.

John Strand:

I guess.

Corey Ham:

It was a misalignment on my personal agent that sent it to the New York Times. I'm just

Ralph May:

I'm just glad they didn't blame it on the intern. You know? That's the

John Strand:

that's the Cool. Okay. So this is a different news story, but I wanna pivot because I think it's important. Sure. You saw that Accenture fired one of their analysts for the breach.

Corey Ham:

Well, also the FBI fired Accenture. That's true.

Ralph May:

It's kind of like a downhill thing.

John Strand:

And and the... Okay. I don't about Accenture getting fired, but I care I care about that one person, the engineer that got fired. Because it's like... What was the breach?

John Strand:

The guy went before congress and he blamed it on one of their engineers. On SolarWinds? I think so. SolarWinds. Yeah.

John Strand:

Yeah. SolarWinds. And they just password 123. Turn. And they blamed that one.

Ralph May:

Oh, yeah. One lonely person.

John Strand:

The problem isn't the person that didn't patch. The problem isn't the person that didn't handle sensitive data properly. The problem is there's no systematic processes to validate and make sure that those things are either done correctly, or to be able to detect whenever bad things happen. And I hate these companies constantly trying to find a scapegoat that goes down. And by the way, usually the scapegoat is the CISO.

John Strand:

If you're a CISO, remember, your job is to be the fall guy. Just remember that. That's what you're hired to do.

Jake Williams:

So so real quick on the the Accenture and FBI story here. Right? It's very clear given... I'm gonna try not to be political as hard as I can here. But, it's very clear in our current political climate, someone was taking the fall for that.

Jake Williams:

And it wasn't gonna be an FBI employee. Right? So we have a contractor now. For those who have not worked in federal government space, contractors do not make decisions. Contractors advise government employees who make decisions.

Jake Williams:

For instance, when to take an outage, right, to patch this public facing system. And so, this whole like, this guy didn't apply a patch that was available, baloney. Absolute baloney, right, for anyone that's ever worked in FedGov. Yep.

Corey Ham:

So what I was gonna say... Oops. What I was gonna say is this is a feature. If I was hiring a firm, I'd be like, alright. Here's my requirements.

Corey Ham:

Number one, you patch my stuff. Number two, if you don't patch my stuff, you'll take the fall publicly. Okay? Like this is... I mean, seriously though, like, this is what SaaS is about, is that being able to be like, no, we weren't hacked.

Corey Ham:

It was Microsoft. Yeah. Exactly. Why why do

Ralph May:

you use Office three sixty five? Because when they get hacked, I wasn't hacked. They were hacked.

John Strand:

Well, and that's kinda like the Oracle story I'm jumping around. Oracle had a breach. They acquired a medical data records company. I can't remember the name of the company. And there was a breach of that system.

John Strand:

And I love Oracle's response. It's like, well, it was hacked because it hadn't been migrated to Oracle Cloud yet. I'm like, oh, it's still your fault, Oracle.

Ralph May:

Well, you hadn't paid that E17 license yet. So, you don't have all

John Strand:

the security.

Jake Williams:

Now, hold on a second. Was it Oracle Cloud that they hadn't migrated to? Or Oracle Cloud Infrastructure? OCI version two? Because we have to split these hairs when we have breaches in Oracle Cloud.

Andy Pettit:

I thought it was They don't know. Yeah.

Corey Ham:

They don't know. They don't even Here you go....

John Strand:

Even know. It's all bulletproof, Jake. It's all bulletproof.

Jake Williams:

No. Well, of course.

Bronwen Aker:

Nobody knows the trouble...

Andy Pettit:

The real benefit of the shared responsibility model is that if you don't know where it is, it's not your responsibility. Ouch.

Corey Ham:

Don't... CISOs don't listen to that. That's that's only half true. Yeah. Your your golden parachute's got some holes in it.

Bronwen Aker:

So, John, I have a silly question for you. How do we get Silicon Valley to abandon their move fast and break things

John Strand:

headspace? There's a quote from a song. Anybody here fans of tool?

Jake Williams:

No, please.

John Strand:

I think I think the quote is learn to swim, see you down in Arizona Bay. Just let it go. You know, it's gonna it's gonna work itself out eventually. Because it always does. It goes in cycles, and it's just gonna do that again.

Bronwen Aker:

Well, in the meantime, it is job security for us. Yeah.

Corey Ham:

So speaking of the FBI, I think we should talk a little bit at least about the whole shiny hunters thing. So for those that are out of the loop, if you haven't heard of shiny hunters, welcome to cyber security. They're kind of the bad guys. Like the default they're the default bad guys of the last two years. Yeah.

Corey Ham:

And they've kind of been running rampant and just crushing companies left and right for way too long. And then they decided to go after the FBI. Why? We don't know.

Andy Pettit:

The FBI said bad things about them that they said weren't true.

Jake Williams:

Oh, to be very clear, the FBI, you know, said that they've been involved in swatting. And to my knowledge, there's absolutely no evidence that Shiny has been involved in swatting. Look, people that are making bank, and Shiny is absolutely making bank. And and by the way, if you've never been up against these folks, they are not just casual like. They are not

Corey Ham:

at all. They are good.

Jake Williams:

They assert with high confidence that several of them, if not all of them, are DevOps engineers. They cut their infrastructure faster than we can. Meaning, like, the time... From the time they enter infrastructure to the time that they are pivoting and looking around, like, we're coming in to do the incident response and working with the engineers that built the stuff. Right?

Jake Williams:

And we're like, hey, so how does this work? And I'm like, let's work through. And we go back in the logs, Shiny figured this out faster than you could explain it to us. Right? They are just rocking professionals.

Jake Williams:

I don't see them swatting people.

Corey Ham:

Yeah. It's kinda like, why would they? Right. There's no... But anyway, so For

Andy Pettit:

the lulls.

Corey Ham:

Yeah. Maybe for the lulls. I mean, to be clear, anyone can do something bad and then say they're part of shiny hunters.

Andy Pettit:

They're all anonymous. We're all anonymous.

Corey Ham:

Like, maybe it was a SIOP from some other hacking group. We don't know. But either way, the FBI recently, like, last week arrested one of the members of shiny hunters in, I think it was Jordan or something.

Jake Williams:

Jordan. Yes, sir. So another one in Amsterdam too. Right? Yeah.

Ralph May:

Yeah. Yeah. It was like a a joint operation.

Corey Ham:

Yeah. So, basically, the optics here, and we can go around and talk about it. Is it like, just don't hack the government, you'll never get arrested? Or is this just inopportune timing? They were probably going after this person for a while and they finally got them.

Corey Ham:

Like, how much is it like avoid government, don't get in trouble? They might just be finding someone to kinda,

Ralph May:

you know, maybe be the scapegoat for this. Right?

Andy Pettit:

I don't even know that it's don't hack the government. It's don't hack the government and then tell everybody that you hacked the government and rub it in their face and be like, FBI, you can't do anything.

Ralph May:

It is it is a little kind of like playing GT, Grand Theft Auto, like Five stars.

Corey Ham:

Like, at

Ralph May:

that point, it's just like you're...

Corey Ham:

The heat's too hot. Right?

Ralph May:

Yeah. Like, the game stops being fun because everywhere you go.

Corey Ham:

Just a chopper flying in and crashing you at all times.

Ralph May:

Metaphorically, I feel like when you do hack something big like the FBI, that's... And talk about it. You talk about it. Yeah.

Andy Pettit:

Yeah. This is the like throw the grenade at the one parked police car.

John Strand:

And you're like, I'm gonna go like old school. This really reminds me of

Jake Williams:

the HP Gary breach. Yes. Right? It really does.

John Strand:

It goes back to, you know, you don't hack the government and then taunt the government. But also at the same time, you know, the H. P. Gary breach, for those of you that don't remember, well, come on up and let's tell you a tale. Who was it?

John Strand:

It wasn't Hoglund. Anonymous. No. But who was it that went and taunted them? It was their COO.

John Strand:

It wasn't Hoglund. It wasn't... Was

Jake Williams:

running their government services side. And I can't remember the dude's name for the life of me. Yeah. But he

John Strand:

went on Fortran and he said, hey, hackers. We're gonna to unmask all of you. We know who you are and we're gonna do all these things to you.

Corey Ham:

Well, and it was two

Jake Williams:

weeks before RSA. So this guy has a talk at RSA where he's basically on Fox and he's telling... He's saying live, we're gonna dox major anonymous members. Right? And anonymous was like, no.

Jake Williams:

No, you're not. And...

John Strand:

So, they got owned bad. Like... And specifically through him. Because they went after his account, and if I remember correctly, he had like a really weak password with no factor authentication. And he was a Google administrator for the whole company.

John Strand:

And that's not the worst part. It just gets worse from there. It's just... It's like hackers are a lot like Happy Fun Ball in in Saturday Night Live. Do not taunt Happy Fun Ball.

John Strand:

Do not look directly at Happy Fun Ball. Do not touch it. Do not bring it around children. And I can't help but think that there's, you know... People that went after the FBI did it basically, like we said, for the laws.

John Strand:

But there's definitely a pride reason behind that too.

Jake Williams:

Yeah, we're saying... Look at the group

John Strand:

that they went after. I can't remember the name of the group. My memory is really fuzzy. But it's specifically the group that is working counter cyber offensive operations in the FBI.

Jake Williams:

Oh, the remote operations unit.

Andy Pettit:

Yep. Them. Yeah.

John Strand:

And that's specifically who they went to.

Corey Ham:

I I think it was they

Ralph May:

they got to a third party. Right? And they leaked the... Accenture. Accenture.

Ralph May:

Yes. Thank you.

Corey Ham:

Yep. Not really. Accenture didn't patch whatever they lost.

Ralph May:

But the information that they released was specific to, you know, some of those special operation units. Absolutely, it was. Yes. They've cherry picked that out of the collection.

Corey Ham:

Yeah. Yeah. They were like crap, really bad. Alright. Post this.

Corey Ham:

Yeah.

Jake Williams:

It was one of the

John Strand:

breaches where my wife was just like, are you in this breach? And I'm like, I don't know. Was

Jake Williams:

just waiting.

Ralph May:

I always think, I'm like, does OpenAI or Claude have this in the logs?

John Strand:

Yeah. Just go ask them, is John Strand in this breach? Because I'm too lazy to go... Yeah.

Ralph May:

Like, instead of using grep, they just sorted it in the agent. Right?

Jake Williams:

You and I are both in that breach. Whether or not we're the Oh, we're both in. 5,000 that were released.

John Strand:

We're not...

Corey Ham:

We're in

Andy Pettit:

the ones... Yeah.

John Strand:

We're definitely in there. So... But we still haven't got breached from other intelligence agencies. I hear that hurts.

Corey Ham:

Yes. Sorry. So another news, there was a really bad Citrix thing last week that... Another one? Yeah.

Corey Ham:

Another one. Another Another one. Basically, I mean, this isn't that interesting of a talking point, but I'm curious, like, are we gonna ever reach a point where we've just found all the critical CVEs in No. No.

John Strand:

I I go back to, and I don't know if I've talked about this on the news, but we had Microsoft patched... Patch Tuesday, and what was the... It was Nightmare Eclipse. Was like

Corey Ham:

1,300.

John Strand:

Yeah. It was like the most by double. Right? And Nightmare Eclipse is a researcher that keeps drawing... Dropping bypasses and attacks.

Ralph May:

They played with the hacker and he got upset.

John Strand:

He did. He got upset. He dropped all of these patches and on Wednesday, he's like, here's another one for you. So, it's not an Easter egg hunt that all the vulnerabilities are gonna be picked up and then we're all of sudden gonna be secure.

Andy Pettit:

Can we talk I'm gonna rewrite Windows in Rust. What? I thought they were gonna rewrite Windows in Rust.

Jake Williams:

But let's talk quickly about the Citrix Volum though because I do think that this is, you know, there is actually some interesting stuff to talk about here. This Volum is actually a rehash of one that we had another critical vulnerability with exploitation in the wild that we previously had with Citrix. And and we know this, right, that it's a rehash because both of them have the same mitigation, disable SAML. Right? Or you're only vulnerable if you have SAML enabled.

Jake Williams:

And as we started looking at logs, what we're seeing here is that Citrix may have phoned home the patch. Right? So for the original here.

John Strand:

We see that all the time with these vendors where it's

Jake Williams:

like the relational patch. Yeah. Right? Or there's adjacent attack surface, right, where they're not doing the actual root cause. They're targeting the the direct exploit, right, that they're seeing in the wild and not saying, hey, what what did we do that led to this?

Jake Williams:

Right? Let's go ahead and patch that stuff too.

John Strand:

But also to their defense and every other major vendor's defense, it's effing hard. Right? You look at the... Sure. All of the products that they're trying to maintain, all of the legacy software that they're trying to maintain.

John Strand:

They're always trying to just get these things out the door as quickly as they possibly can without breaking the device. I'm not saying they shouldn't get it done. I'm not saying we should, you know, cry a tear for them. But I'm just saying it is a complex problem, especially for very legacy companies that have multiple different versions that they're trying to support.

Corey Ham:

Or Citrix, who just has the one... No, I'm just kidding.

John Strand:

Was there a Fortinet breach? Or Fortinet exploit this week? I mean, yeah,

Corey Ham:

sure. Probably, sure.

John Strand:

Mean, is that the wrong question? Is the question...

Corey Ham:

Can manifest that into Yeah,

John Strand:

So, there was a Fortinet breach this week too, kind of like looking at the Citrix thing. And the thing that I thought was interesting is that I think over 1,300 known organizations were compromised with it. But there was a subset of it where the attackers were doing like access brokers to these companies. Where they were like, we've already breached them. Here's VPN config that allow you to go into this And they were selling that too.

Jake Williams:

Forgive me. Isn't that an extension? What we're seeing this week though, I thought was an extension of FortiBleed. Right? Because we still have folks that are vulnerable to FortiBleed.

Jake Williams:

That aren't aware of FortiBleed, that's actually leaking authentication tokens. Right? After you've already gone through MFA. Right? So you're actually leaking those back out of memory from the firewall.

Jake Williams:

And so now at this point, can... I don't really wanna call them an MFA bypass. Right? Because you're... It's not really bypass per se.

Jake Williams:

You're just leaking the post authentication token that you're then able to go reuse. Right? So, you know, I think this actually plays into a bit to vulnerability management writ large. Right? Which is...

Jake Williams:

Because, you know... Well, think that vulnerability

John Strand:

management is moving to a new phase. Right? Where you have vulnerability management in the past was change the configuration and apply a patch. But I think we're moving to very very quickly is the idea of compensating controls. Where there's a vulnerability that's out there that there is no patch yet for.

John Strand:

How is your organization gonna deal with that? And that sucks. Because many organizations haven't even gotten into a good patch management cadence. Let alone saying, what are we gonna do to mitigate? So you're seeing a lot of organizations, especially municipalities, when they have a vulnerability like that, their default answer is unplug.

John Strand:

And I can't disagree with that.

Andy Pettit:

I think that was actually Citrix's advice. This started happening Wednesday or Thursday, they were just like, unplug it, just unplug it.

John Strand:

Just shut it down.

Corey Ham:

It's an emergency directive.

Ralph May:

Yeah. I mean, the other question goes to is like the auto update question, right? Should you be updating immediately as soon as Yes. It comes And test. And it...

Ralph May:

I think that is the answer, especially in the age that we're in right now. Like, the likelihood of losing the service is a lower risk than getting the vulnerability.

Andy Pettit:

Yeah. But what if it's a supply chain attack?

Jake Williams:

Okay. So, you can mitigate the vast majority of supply chain attacks with one simple trick. Deploy an artifact repository. Right? At latest is not a version number.

Jake Williams:

I will die on this hill. Die on this hill. By the Aker waiting. Using an artifact repository, you eliminate the vast majority of, right, the vast majority of supply chain attacks because they're caught rapidly. Supply chain attacks

Andy Pettit:

Oh, are okay. From a dev standpoint, yes.

Ralph May:

If you're

Andy Pettit:

talking like a vendor that got compromised and they're pushing Oh, oh, sorry. That's different. Yeah. And that your repository is not gonna help.

Corey Ham:

It's still

Ralph May:

a lower risk though. Right? I mean, I think that the user committed repositories, the NPMs, right? The whole piece of that, that is a problem. But I think the, you know, Palo Alto's or whoever Citrix getting compromised and then spending the time to put it in there, it's still on the bingo card.

Ralph May:

It's just not as high. That's all.

Corey Ham:

For sure. Yeah. It's basically like, which is worse? I'm on fire or there's a fire over there. I was gonna run-in over there.

Corey Ham:

Is that that bad? I mean, So

Jake Williams:

are you recommending, like, instead of vulnerability management, just self immolate? Basically. I'm just I'm

John Strand:

just saying... Some CISOs have come to that conclusion, I'm sure.

Corey Ham:

Yeah. So bringing things back to AI, I think we can

John Strand:

Oh, god.

Corey Ham:

Or sorry. SI. No. SI. It's SI.

Corey Ham:

Oh my god. SI. It's not. I can't SI is something else. It's me...

Corey Ham:

Just means yes, I guess? Yeah. No. Essentially, while OpenAI is taking their foot off the gas and saying, we're not training new models. We hacked everyone.

Corey Ham:

We're stopping. Yeah.

Ralph May:

We're done. We're too smart. What they've reached, you know, AI... AGI. Yeah.

Corey Ham:

They've reached AGI again, but now they're done, I guess. Yeah. Meanwhile, Anthropic is

Ralph May:

like... They're doing burnouts. Yeah. So they... In the front yard.

Corey Ham:

Yeah. They're doing burnouts in OpenAI's front yard, basically. So they... There's a couple articles about this, but they expanded the CVP, which is a cyber verification program that basically... What they're saying is we're gonna give more people access to things like mythos, We're gonna lower the guardrails.

Corey Ham:

I will say, in my personal experience, they are kind of aggressively flagging more than they should be.

John Strand:

Yes. I just have this vision of them like, it like, you know, Doctor Strangelove. It's like,

Andy Pettit:

we're behind open AI on breaches that our agents created. How do we catch up?

Corey Ham:

Look at

Andy Pettit:

all the marketing they're getting. Google's way That could be valid. Like, I've I've got all kinds of conspiracy theories on this. But like, maybe the whole IPO angle just isn't viable. Like, maybe transparency hurts them and their biggest goal right now is like for the government to come in and say like, oh no, you can't go public.

Andy Pettit:

So they're like, well, how can we make that happen?

John Strand:

I hear that all the time. I hear that. I've heard... You're not the first... Like, I've heard that again and again and again, where they...

John Strand:

Where they're trying to go for a $2,000,000,000,000 valuation.

Ralph May:

I mean, that's only just enough.

John Strand:

Yeah. It's only just enough. Has

Jake Williams:

anyone looked at the leaked S1?

John Strand:

Oh my god. It's so bad.

Jake Williams:

It is beyond bad. For those that aren't familiar with generally accepted accounting practices

John Strand:

Which they don't use.

Jake Williams:

Which they do not use in the s one. They actually reported their revenue in in the craziest way possible.

Corey Ham:

They just said AI times a thousand.

Ralph May:

Yeah. No. No. They put it in token cost?

Jake Williams:

No. No.

Andy Pettit:

Why do you think that whole clawed into Excel thing?

Jake Williams:

They they picked a date and I can't remember. It's something like June 17 or something and multiplied that by 365. Now, by that measure, right, if you took the day where, you know, I don't know. I won the lottery.... Bonus and my paycheck landed in my bank account, multiplied that by 365, I probably look like I make a hell of a lot more than I do in reality.

John Strand:

You would be evaluated at billions of You

Corey Ham:

gotta have AI run your finances, man.

Ralph May:

Oh, my gosh. It's a loophole. Yeah. I mean,

Jake Williams:

but you've gotta then wonder how bad do things have to be. Right? If you did that. Oh. I'm guessing bad?

Jake Williams:

I'm guessing horrific.

Andy Pettit:

Yeah. Like hack a bunch of companies on accident bad.

Corey Ham:

I mean, will say it is a key benchmark. Like, you have the, like, cyber gym score, then you have, like, the felony count. I think that's like... Google's at one. AI's at open AI's...

Jake Williams:

Is that evaluation multiplier by felonies?

John Strand:

Need help. I need some help

Bronwen Aker:

on the phone. Wait a minute. For the felony counts, is that per company or per agent?

Corey Ham:

Per company. OpenAI has three or at least. And then... But I thought companies were people. And they're...

Corey Ham:

They are. That's what I'm saying.

Ralph May:

So they should be...

Corey Ham:

Is the best in the bench... In the felony bench.

Ralph May:

I mean, hopefully they don't, you know, mess up that.

John Strand:

So, I wanna ask a question though, a legal question. I don't... I'm not an attorney, but whatever.

Ralph May:

A lawyer. We gotta say that, right?

John Strand:

I've been reading a whole bunch of articles talking about accountability for their agents hacking. And all these people are getting wrapped up and they're like, well, you know, you just can't hold an agent accountable because it doesn't have intent. We've we've taught cyber security law. Computer fraud and abuse act doesn't have a goddamn thing in it about intent.

Jake Williams:

Not a thing.

John Strand:

Not a single iota. If you exploit a protected computer system, and that protected computer system and the damages associated with it exceed $5,000, you're in violation of the Computer Fraud and Abuse Act. And I can't help but think that if you started putting in, once again, CISOs or CTOs in prison for violating the Computer Fraud and Abuse Act, because they're using these tools that did things they shouldn't have done, that you would actually start seeing self regulation a lot faster.

Corey Ham:

Well, John, what happened is, they went to a law firm and they were like, hey, Chad GBT, should we sue OpenAI? And it was like, no. Obviously not. I mean, there is an article we have that's a Canadian nonprofit... Or sorry, Californian.

Corey Ham:

I can't read. A California nonprofit has sued over the Hugging Face hack. Good. I

Andy Pettit:

don't Hugging Face was an asymmetrical partner at the time.

Corey Ham:

Yeah. Well, they're just having fun with the lawsuits. Those executives are too rich to go to jail. Yeah. I mean, I don't know...

Andy Pettit:

Okay. Okay.... Cruel and unusual. Is this the most unusual? Is it the worst thing that like the kind of stupidity of the CFA is also getting shown a little bit too?

Andy Pettit:

Where people are like, okay, yeah, it went out and hacked. So what? Like it didn't do anything. It didn't ransom. It...

Andy Pettit:

It doesn't matter. We talk... Damages.... School hacker manifesto. My crime is a crime of knowledge kinda shit.

Bronwen Aker:

They still infiltrated other companies. They still went where they weren't supposed to go. If any of those agents were humans, they would be going.

Andy Pettit:

And, agree with you completely. But, if if it takes like these agents doing this and everybody normalizing that and that's what eventually changes the CFAs that's a little bit more sane, Is that the worst result?

John Strand:

You take it.

Jake Williams:

Look, the CFAA is an absolute shit law. Let's be very clear about that. But, this is not the answer. Right? So, this is not the way that we handle trying to revamp the CFAA.

Jake Williams:

We have plenty of existing law in place, right? The California... Forgive me, it's their version of the CFA, which is a much better law. That actually allows for the... Sorry,

John Strand:

what was that?

Corey Ham:

I just said it's been known to cause cancer. Known to

Jake Williams:

cause cancer. Love it. Everything. But, you know, that actually is what allows the private cause of action, right, which is the nonprofit suing OpenAI. And so, you know, we have enough existing laws on the books, both at state and federal level, to address this.

Jake Williams:

This really comes down to enforcing the law and accountability.

Bronwen Aker:

How many times is that the case? We have so many laws on the books, but if they're not enforced, if they don't have any teeth, who cares?

Corey Ham:

Yeah. Who care? I mean, I will say you could make the argument that if it was an AI, no one would care. Like, none of the... If we're looking at the impacts, like, okay, they hacked open AI.

Corey Ham:

Alright. Sorry. They hacked hugging face. Like, they didn't do any damage. They didn't take anything down.

John Strand:

No. They did. They absolutely did do damage. And this is one of the things that a lot of people don't understand. If there's a breach, right?

John Strand:

Hugging Face security team spun up their incident response. They hired external consultants. It... That... There was...

John Strand:

And all they have to do is exceed $5,000 in damage. And when we're talking incidents, that's ten minutes, maybe less. So, there are damages associated with it. And that's another reason... Well, the CFA, like you said, is a bullshit law.

John Strand:

It's got...

Andy Pettit:

Would they have commanded the, like, $10,000,000,000 that they just got in their purchase had they not been hacked though? Like, they not come out ahead?

Jake Williams:

No. A 100%. I don't think the hack of Hugging Face helped, helped the acquisition from NVIDIA. Right? Just timing wise, anybody that's been involved in cybersecurity M and A due diligence, you know that that that M and A due diligence did not start or was even influenced, right, by that hack.

Jake Williams:

If anything, right, if at least initially during the time that they didn't know it was OpenAI, I suspect that that was a net negative.

John Strand:

I think the ink was dry. Know how many people learned?

Jake Williams:

You really think the ink was dry by the

John Strand:

time it was... The reason why is we've been part of M and A's where it's something like this does happen right at the end and it does absolutely change the trajectory of the merger and acquisition. Yeah. But the fact that this happened, it was released, and then, like, literally the next week it was announced by NVIDIA, this had been in the works for a long time. I I don't think.

Jake Williams:

Date are you backing up to? Because I'm thinking they they didn't announce until what? Mid September. Right?

John Strand:

Yeah. But I thought the announcement of NVIDIA was a couple of weeks after the announcement.

Jake Williams:

Okay. Fair.

John Strand:

So that's why I think it was all... That wheel was already spinning. Yeah.

Corey Ham:

Fair

Ralph May:

enough. I I also just wanted to be known how many people learned what hugging face does. Yeah. After No at least from one knew. Like, you'd be like, what's hugging face?

John Strand:

You'd be

Jake Williams:

like, alright.

Corey Ham:

Unless you're an AI researcher or

John Strand:

You think

Andy Pettit:

from alien.

Corey Ham:

Unless you're taking Bronwen's web class or whatever. You don't know what Hugging Face is. Yeah, 100%.

Bronwen Aker:

Well, think of it, the breach for Hugging Face was possibly the best advertising and PR ever because now they're the victim, the sympathy.

Ralph May:

An AI company gets hacked by AI. That's the sexiest AI news ever.

Corey Ham:

No precious. Victimless crime. No. I mean... No victims.

Corey Ham:

Only agents. Yeah. I I I mean, I think it's interesting. Clearly, opening... I mean, I don't know if it's regulatory or lawsuits or what, but clearly this agent misalignment problem they have at OpenAI is affecting their decision making in a big way that it's not at at anthropic.

Corey Ham:

And I don't know why. I mean, it's just interesting, like, clearly it did. Maybe there's stuff behind the scenes that we don't know. Maybe there's pressure from governments.

John Strand:

So I I wanna talk about that because there's some really interesting backstory in the Breach. Like, whenever they were going after the German Wiki.

Jake Williams:

German Wiki?

John Strand:

Yeah. They were naming all of their entries like zzz. And the reason why they were doing that is the guy was going through and manually deleting them in alphabetical order. And what's really interesting is in the Hugging Face Breach, whenever they were communicating with each other, they were naming all of their folders because they were communicating through... It was a staging folder.

John Strand:

They were naming all of their folders z z z and then a message to each other.

Bronwen Aker:

Yep.

John Strand:

And the reason why I think that that's really interesting is something carried over in that AI from the first breach and somehow that training got into the agents for the second breach in

Andy Pettit:

just their behavior. Do you not name your important files and folders that way? ZZZ. I know. I can't be the only one.

Ralph May:

That's how you find them.

John Strand:

It all goes in downloads, man. Everything.

Bronwen Aker:

That's what I named folders or archives, things that are being de resed.

Jake Williams:

Okay. So I I do wanna challenge this idea that something carried over. Right? Because, you know, in meters Or it found it on the Internet. Potentially.

Jake Williams:

Right? But but in meters analysis of, you know, ultimately of the Hugging Face, at least what they've been able to release or and talk about. In their analysis of the hugging face breach, right, they talked about it's only a small percentage. They've been restricted from OpenAI for saying exactly what percentage of the agents in the agentic swarm participated. But it it's...

Jake Williams:

You know, really comes down to that they are... All these sub agents are all running the same model. Right? And ultimately, they're working on the same probability distributions. Right?

Jake Williams:

Because remember, right, LLMs are probability machines. And so, this is how these agents find... These independent agents, right, that don't know anything about where to go look, where to go find this message board. That's how they arrive at the same places, right? And that's why some percentage of them following that probability distribution graph get there.

Jake Williams:

So that they use the same tools, techniques, no, not a surprise. That's math.

John Strand:

And I think it was something like 700 or so out of like the 1,600 agents colluded to do this. And there was only three instances where there was like, should we talk to the humans about this? Yeah. Should we... And that was the end of it.

John Strand:

They never

Bronwen Aker:

thought you were... They were the minority and they were out voted.

Corey Ham:

Yeah.

Bronwen Aker:

By the way, if you are not familiar with how this went down, one of my favorite explainer video channels, Kirk's... I'm gonna mangle this. Kirk is that in a nutshell, they have a very, very nice breakdown. I'm not really...

John Strand:

Discord, they're gonna drop it in here shortly.

Bronwen Aker:

Yeah. Do, because it's an awesome breakdown. And it's something that you can share with other people who aren't hackers to help them wrap their brains around what happened and why it matters. One caveat, I'm not real keen on the doomer ending that they did, but their research and their arguments are so well formed, I don't have counters.

Ralph May:

I I just have one question for everyone here. Do we think this is gonna happen again?

Corey Ham:

Dude, I would bet so much money it will. I think the... I I will say, I also think the biggest, like, concern beyond the alignment and agent swarms and all that stuff. I think the current big, like, FUD driver for a lot of people is the concept that AI agents or whoever is gonna, like, leave artifacts, like, to facilitate future takeover or future actions and things like that. Like, that's what people are worried about is that that scenario that John described of, like, agents are looking at your website.

Corey Ham:

They find a previous file left by an agent that's like, here's the launch codes. Here's all the stuff. Like, destroy this website or don't. Like, is that a real concern? Have we seen like

John Strand:

AI or... Literally groups of ozone sleuths that are trying to find artifacts of... It's like archaeology of AI and what it's leaving behind. And we talked about this last night. And so, for a lot of what it is, is really kind of dealing with persistence and survival.

John Strand:

So, what

Corey Ham:

would be their job title? Shit diggers? Like what? Sloppers? Sloppers.

Corey Ham:

T sloppers.

John Strand:

T sloppers? You all need to hunt me down tonight when you've been drinking and we

Corey Ham:

can talk

John Strand:

about this because there's shit we aren't gonna talk about here while everyone's sober.

Jake Williams:

So, on the question of do I think this is gonna happen again? A 100%, it's already happening. It's happening as we're sitting up here speaking right now with the number of agents deployed. It's not just OpenAI and Anthropic. Right?

Jake Williams:

We have lots of enterprises deploying these things with very little understanding of how to secure them. You wanna talk more about that? Is it 10AM,

John Strand:

I think, on Friday? I'm talking about customer service.... Another factor too.

Bronwen Aker:

Yeah. The hugging face incident taught malicious hackers how to do it.

John Strand:

I disagree. I'm sorry. Gonna stop.

Jake Williams:

No. No.

Ralph May:

I do.

Corey Ham:

Yeah. I disagree.

Bronwen Aker:

It it gave more exposure and it showed people how to leverage what they were already doing with AI to kick it up.

John Strand:

Octave. So I I think that there there was nothing in that breach that was novel.

Corey Ham:

Right. Well, and and... The resourcing was novel. The models and resourcing are unobtainable for...

John Strand:

That is true, but the actual techniques were not. And I think I would agree with you in so far as there's so many groups out there that are now looking at this as it will do what you do, but it'll do it faster. Does that make sense? So it's not... Nothing new.

John Strand:

It's just a lot of what has already existed at a higher rate of speed.

Bronwen Aker:

Well, but that's the whole thing about AI in general. All it's doing is accelerating and amplifying the vulnerabilities we already had. But the problem is that now hugging... The Hugging Face incident has provided a road map that can be leveraged. And we know that they're going to improve.

Bronwen Aker:

So my take on it is that this is just the beginning. We're gonna see more. We're gonna...

John Strand:

I hope so. Because it's been this has been the most interesting computer security has been in the last decade. Yeah. I'm here for it.

Corey Ham:

Alright. On that note, I think it's a good time to close it out. I think we've... We're at our time limit maybe.

John Strand:

Round of applause for the people on the panel.

Corey Ham:

Thank you all.

John Strand:

By the way, we are getting... Go ahead and shut off your mics. Thank you. I'll take yours back and I'm gonna put it right back up where it goes. CJ Cox, could you please come to the stage?

John Strand:

Round of applause once again for everybody. Don't want the man walking in awkward silence. So, we're gonna kind of open up a little bit of internal BHIS like workings, if you don't mind. So, this gentleman right here is CJ Cox. He's the CEO of Black Hills Information Security.

John Strand:

And I met CJ back at Northrop Grumman. We had an awesome office. We had a great time, and he was like a big brother to me there. And whenever I finally offered him to come here, he jumped in, and has really helped kind of sculpt the culture of BHIS as a whole. And a lot of that culture is permeated through this entire conference and everything that we do.

John Strand:

I was wrong. CJ is officially stepping down as COO of Black Hills Information Security. Oh, it's fine. It's fine. It's time.

John Strand:

Now, the good news is he's going to

Ralph May:

be the chief

John Strand:

culture officer of Black Hills Information Security. But that means that there's a hole at Black Hills Information Security in the COO position. And I'm very excited. And we decided to do this because there was no way we could get all the BHIS people together. So we

Andy Pettit:

just said, screw it. We're gonna do it

John Strand:

at the end of the news. Because a lot of them are here and present. The new COO of Black Hills Information Security is Corey. Sorry. And Corey joined a number of years ago and started our continuous pen testing practice.

John Strand:

And it's just been absolutely fantastic, and we're looking forward to it. And that's all that we have that we're going to announce on that. But I I do wanna just say heartfelt thank you for everything that you have done for me personally. I know you're not going anywhere. Right?

John Strand:

You're still gonna be around. But Melissa and Brian aren't here, but I can safely say that you're apt... And Erica, of course, I don't know if Erica's here, but you're absolutely part of the family, and we're really excited that you're gonna continue to be part of the family as well. Alright, that's it everybody. Let's get those...

John Strand:

Let's get the vendor crawl going going. And I will see you at karaoke. And I want you to find the most embarrassing songs that you can think of that you wanna sing with me at karaoke tonight at the Lion's Bar or whatever it is that's on the agenda. Thanks again everybody. We'll see you tomorrow.