Know what you have online—and what your AI assistant can reach.
Sources & transcript: Episode #56 page
Read transcript: https://share.transistor.fm/s/891bfa6c/transcript
Apple Podcasts web strips note links: use its existing Episode Webpage action for clickable sources and the transcript.
Automated transcript; may contain errors.
Know what you have online—and what your AI assistant can reach.
Shelby Pettit and Bryce Kunz discuss Daniel Miessler's three recommendations for “The Hackening”: inventory your online assets, reduce unnecessary internet exposure, and reassess small security vendors holding sensitive credentials. They cover practical asset registers, AI-assisted discovery, management-connection logs, Tailscale and Cloudflare Tunnels. Bryce challenges the vendor-size recommendation: bigger is not automatically safer, and security choices still need to fit your people, business strategy and risk appetite.[1]
Then they unpack Profero's Grok Bot security assessment. Starting from one assumed hostile command inside an agent, researchers reported a path through a sandbox-reachable Docker API to that account's VM, exposed credentials and persistent storage. The discussion explores agent permissions, credential lifetime and limiting the blast radius of personal AI assistants.[2]
Important scope: Profero did not demonstrate an escape into the underlying EC2 host or another customer's VM. The reported “local” token belongs to the sandbox exec daemon—not a customer's Grok login—and remote access also requires an account-specific routing token. This episode is a discussion of reported research, not an independently reproduced exploit or a report of a confirmed customer breach.[2]
Source note: Profero's November 8 report said the findings remained unfixed. The on-air expectation of a vendor fix is not confirmation that one shipped; later remediation has not been independently verified here. Profero recommends dedicated agent identities, narrow connectors, and keeping production browser sessions and long-lived keys out of the sandbox.[2]
Plus: Shelby's Magic Maze recommendation and Bryce's Godzilla review.
KEY MOMENTS
0:00 Welcome, hype & getting older
1:10 Three security priorities for the hackening
1:48 Inventory your internet-facing assets
3:23 AI-assisted discovery & management-connection logs
4:20 Reduce unnecessary internet exposure
5:29 The small-vendor recommendation
6:20 Why bigger does not automatically mean safer
8:00 Security budgets, business strategy & risk appetite
9:43 Personal AI assistants & assumed compromise
10:51 Docker-daemon access & sandbox boundaries
12:35 Exposed secrets & persistent storage access
13:53 Agent blast radius & mission-critical data
15:08 Static passwords & security takeaways
15:55 Magic Maze: a silent cooperative heist
17:11 Bryce's Godzilla review
18:27 Wrapping up
SOURCES DISCUSSED
[1] Daniel Miessler — The Post-AI Security Floor (November 1, 2026)
https://danielmiessler.com/blog/post-ai-security-floor
[2] Profero / Omri Segev Moyal — The Rootkit Built Into Grok Bot (November 8, 2026)
https://profero.io/blog/the-rootkit-built-into-grok-bot
Which matters most in your environment: asset visibility, vendor access, or the reach of your AI assistants? Share your take and subscribe for more AI and cybersecurity discussions.
Find the show: https://aicyber.land
Topics: #AI #Cybersecurity #AIAgents #Grok
Join industry experts and thought leaders as we dive deep into how artificial intelligence is transforming cybersecurity, shaping defense strategies, and creating new opportunities in the digital landscape.
AiCyber.Land #56 — transcript
Hosts: Shelby Pettit and Bryce Kunz
Source project label: November 9, 2026 (not independently verified recording start); final production candidate 18:48.29.
AUTOMATED TRANSCRIPT — LIGHTLY REVIEWED; NOT WORD-PERFECT OR HUMAN-LISTEN CERTIFIED.
Generated by Whisper medium from the preserved normal-speed recording, then mapped once to the final video timeline (7.801905s intro + source time / 1.2). The complete transcript was read for editorial review. Only documented certain spellings were corrected; uncertain passages remain marked. Square-bracketed Editorial notes are review annotations, NOT spoken words. Statements in the conversation can be opinions, predictions or inaccurate summaries; the source notes clarify research scope without rewriting the audio. No speaker-by-speaker attribution has been invented. This TXT covers the conversation, not the intro/outro music, and is not a timed app-caption track.
Sources checked:
Daniel Miessler — The Post-AI Security Floor (November 1, 2026)
https://danielmiessler.com/blog/post-ai-security-floor
Profero / Omri Segev Moyal — The Rootkit Built Into Grok Bot (November 8, 2026)
https://profero.io/blog/the-rootkit-built-into-grok-bot
The latter assumes one hostile command in an agent and reports account-VM scope, not a cross-customer/underlying-EC2-host escape. Later remediation has not been independently confirmed here.
TRANSCRIPT — timestamps match the final video
[0:08] Hey, welcome back to the AI Cyber Land podcast where artificial super
[0:13] intelligence meets cybersecurity. And we got the world's greatest co-host here,
[0:18] Shelby! We're your eyes and ears for all things AI and cybersecurity related, the
[0:23] breaches, the breakthroughs, and everything else in between. So kick back,
[0:26] relax, and SMASH that subscribe button! I did it again! Maybe this time do some
[0:31] hype. I still haven't googled what hype is but someday I'll figure that out.
[0:34] That's how you... doesn't really tell my... showing my age here, Shelby. By the way, I
[0:38] went to tacos on Friday and someone was like, how old are you? And I told them my
[0:42] real age and they were like, whoa! I had no idea. Like, I thought you were way
[0:45] younger than that. And I was like, I don't know. Is that a compliment or is that an insult?
[0:49] I was trying to figure it out afterwards. I was like, I don't know. They're like, oh you're so
[0:53] mature! No, just kidding. I know, exactly! Like, you're like a 12 year old boy in your maturity level. So, whatever, man. I like zombies.
[1:02] Your shirt's cool. Yeah, I like zombies in Star Wars. What could I say? What did you see this week, Shelby, that you thought was interesting?
[1:10] So, Daniel Miessler published a blog post and there's some good advice. He gave
[1:15] three tips on what we can do right now to kind of like, what to lock down first now that we're living in the
[1:23] hackening. So, it's no secret that like, attackers are using AI to scan everything, to profile everything.
[1:30] And they've got this nasty combo of like, being extremely persistent and crazy competent.
[1:35] They're like, they can write all sorts of, you know, exploits and things like that. So, we're really trying to step up the game.
[1:41] And I like step one because it also echoes a sentiment I've been hearing from other people, which is tip number one.
[1:48] Make a full list of everything you have shipped online. It's obvious and it's boring, but asset management is, it's a struggle.
[1:55] I don't think I've heard from anyone that was like, yeah, our asset management is always accurate and reliable, right?
[1:59] Like, everywhere you go, people are struggling with that. But it's really hard to defend what you don't know you have.
[2:04] If you don't know it's yours, you know, the attackers are going to find it first. They probably already have.
[2:09] So, we're talking about websites, APIs, apps, databases, containers, anything that's accessible.
[2:15] You've got to figure it out. And if it's too intimidating to like, start with a database to track all of your asset management,
[2:22] you don't have to get fancy. Like, even a markdown file or a spreadsheet is a good starting place because what you need is just the data, right?
[2:30] So, he gives some bullet points on what to grab, info to grab on each asset.
[2:35] You need its name, its address of what it answers to, what it is categorically, where it runs, what data it holds, who owns that data.
[2:44] Is it on the Internet or is it supposed to be accessible only to certain networks or things like that?
[2:50] Which vendors have access to it? Who has credentials to this?
[2:53] And another one that I thought was kind of funny is a timestamp or I guess a date when you last checked that it exists.
[3:01] So, yeah, those are some things you should know about your assets so that if anything happens, you know who to contact and what's at risk there.
[3:08] And you can really do so much more. You know, it seems like everyone wants some shiny new AI tool to help them do XYZ.
[3:14] But if you don't have your asset management in place first, it's not going to be able to work with the kind of incomplete or garbage data, right?
[3:19] So that's step one. I like that step. I like that step, shall we?
[3:23] I just want to put a pitch in there that you can use like your cloud and your ChatGPTs to help like map your networks, right? [Editorial note: ASR wording uncertain: cloud may refer to Claude. Original wording retained.]
[3:29] Like mine, I told cloud to do something and it totally found a speaker that was on my network that I didn't even remember existed. [Editorial note: ASR wording uncertain: cloud may refer to Claude. Original wording retained.]
[3:36] And then the other pro tip I have is.
[3:40] If you're at a really large enterprise and you can do like NetFlow logging for like port 22 or port 3389.
[3:49] Try to like log that stuff over time because then you'll be able to see like when you don't know who owns an asset, you'll be able to see what other IP addresses are connecting into it to manage it.
[3:58] And so sometimes you can like, you know, this person manages this asset, but you don't know what this one is.
[4:02] But if you have the logs that show who connected to it to manage it, then you can like kind of correlate that.
[4:06] So there you go. There's my pro tip.
[4:08] All right. That's a great suggestion because it is hard at large enterprises because it's not a one person job, right?
[4:13] Yeah, and you don't have to log everything.
[4:15] Just log those two ports, you know, so cool.
[4:19] That's great advice. Thank you.
[4:20] So once you figured out what you've put out onto the Internet from your company, step number two is stop.
[4:26] It's just remove as much of your exposure as possible from the Internet at large.
[4:31] So I like this quote from Daniel.
[4:33] He said, in the olden days, you could just put everything behind a VPN or SSH or proxy of some sort.
[4:38] Nobody should be trusting anything right now because there's just so much happening in our space.
[4:44] So what was his solution? He gives to he says, try out either Tailscale or cloud for Cloudflare tunnels and still all your stuff behind that, [Editorial note: ASR wording garbled; Tailscale and Cloudflare Tunnels spellings reviewed, surrounding words retained.]
[4:53] because it's just going to give you an extra layer of protection.
[4:54] They're not infallible, of course, because nothing is.
[4:57] However, it's a lot better than just being open.
[5:00] So with Tailscale, what happens is it builds a private network between your own devices.
[5:04] And that way, services answer only to devices that you've added, which gives you a lot more control.
[5:09] And then with the Cloudflare tunnels, the way that works is it connects outward from your server to Cloudflare so that your server,
[5:16] you can shut down every inbound port just to keep things simple, reduce that attack surface.
[5:22] So that's step number two. Number one, figure out what you got on the Internet.
[5:25] Number two, settle that down, reduce that.
[5:29] And then number three, kind of brutal.
[5:32] And my apologies in advance to anyone who is in a small or medium sized vendor space.
[5:37] But according to Daniel Miessler, you should reduce reliance on small security vendors.
[5:43] And his reasoning for that was that they are usually have less of a security team to be able to withstand these,
[5:52] the caliber of attacks and volume of attacks that are happening right now.
[5:56] And yet they look like juicy targets to attackers because they might have a lot of people's credentials because they have all their customers access.
[6:05] So that's what he said. He said, move to big platforms, go through that list from your your asset info where it says which vendors have access to things
[6:11] and kind of evaluate that and see how comfortable you feel with those things and if it matches your risk appetite.
[6:17] So there are your three steps for how to survive the hackening.
[6:20] All right. Well, I'm just going to say I strongly disagree with number three.
[6:24] Like I have worked for I've seen operations of very large vendors that are horrific.
[6:29] Right. I mean, absolutely horrific.
[6:31] So just because a vendor is large doesn't mean that you're reducing your your risk posture.
[6:38] If anything, you might actually be increasing your risk posture.
[6:41] And the reason that I say that is because the reward is so much higher.
[6:46] Right. Like if I'm a cyber criminal and I want to if I'm able to breach AWS, I can breach everybody.
[6:52] You know what I mean? So like that is not only like is it apples to apples.
[6:56] It's like if you're bigger, you got to be way better, too. Right.
[7:00] So I don't know. I I I'm going to just say my personal belief and, you know, people can counter this if they want is.
[7:10] It comes back to the people. Right.
[7:12] Like which group has the most concentration of qualified people. Right.
[7:17] Because those are everything else aside.
[7:20] That's probably going to be the group that's going to be net more secure.
[7:25] Right. So I feel like there's a lot of individual variability.
[7:28] Right. You can have big companies that are making big mistakes.
[7:31] Right. And then I mean, I'm not going to name better names, smart people at these small companies.
[7:35] So I think I don't really.
[7:38] Yeah, I don't really follow step number three from Daniel.
[7:40] But yeah, I you know, you can listen to everyone.
[7:43] I just I. I'm telling you, I've seen some massive companies and there's no talent there.
[7:48] Right. And so their operations are quite horrible.
[7:51] Right. But they have the vendor connections, they have existing contracts with people, so it's easy for people to procure their services.
[7:57] Right. So I I.
[8:00] And also, I got to say, to a certain degree, you get what you pay for in this space of cybersecurity.
[8:05] Like if you're going for the lowest cost provider, regardless of the size of the provider.
[8:09] Yeah, you're you're running you're you're running some risk appetite there.
[8:12] Right. So so I mean, yeah, I mean, I know people don't want to hear that.
[8:16] But like you, you do to a certain degree get what you pay for.
[8:19] Right. Like you've got to. Where I think the sweet spot in the market is personally is.
[8:25] You never want to go with the lowest price.
[8:26] You never really want to go with the highest price because those people are probably just gouging you.
[8:30] Right. You want to go with like somebody who's like about 75 percent.
[8:33] Right. So somewhere where they're like a little bit above average, but they're not like on the extreme side.
[8:37] So I feel like that's like the usually the best just as a rule of thumb, the best thing for the buck.
[8:43] But I mean, I also have like, you know, strong opinions.
[8:45] So I, you know, and my opinions are clearly right.
[8:48] So you should listen to me.
[8:52] There's no one here is to hear your opinions.
[8:54] Right. Yeah, exactly.
[8:55] So anyways, you know, but, you know, if I am very much like everyone should make the decision that's right for themselves
[9:03] and the business they're representing and every business has a different strategy and therefore, you know,
[9:07] what might be my recommendation to one business would might not be my same recommendation to another business,
[9:13] because a lot of that depends on the overall business strategy.
[9:15] And I know people don't want to hear that either.
[9:17] But like because you have to put thought into what your strategy is and where you're going as a company.
[9:23] But, yeah, I mean, you know, like if you, you know, if you serve sodas versus, you know, you know, put people on the moon.
[9:31] Right. You probably need a different risk posture. Right.
[9:34] So. All right. Anyways, that's done.
[9:35] You got me going. You got me fired up, Shelby.
[9:37] Great job. Great job.
[9:39] Tell us what you did in the morning.
[9:40] But you did it.
[9:42] OK, so this is what I saw.
[9:43] That was cool. There's an article that came out this morning about Grok Bot. [Editorial note: The original Profero page displays November 8, 2026; this morning is the host's on-air wording.]
[9:47] And so essentially what the assumption is, for those of you who don't know,
[9:50] Grok Bot is like a personal assistant and you can have like a chief of staff Grok Bot and then you can have multiple sub Grok Bots.
[9:56] And this is honestly, in my opinion, from like a user standpoint, one of the best personal assistants that are out there right now.
[10:05] It just seems like it's the smartest and seems like it's the most efficient.
[10:09] I don't personally use it a ton because I use my own custom setup on using Hermes on hardware I own.
[10:15] But for people that don't have that, I typically recommend they go to Grok Bot.
[10:18] So a competitor Grok Bot is like Meta's Muse or OpenAI's Dot or Clod's coming out. [Editorial note: Competitor/product names in this ASR passage are uncertain and were not independently verified.]
[10:25] Anthropic coming out with one called Duet. [Editorial note: The company spelling Anthropic is corrected; the spoken/ASR product name Duet remains unverified.]
[10:28] So, OK, so they had a customer ask them to take a look at Grok Bot to see how secure or insecure it is.
[10:34] And they started with the assumption that someone was going to be able to trick your bot into running like one malicious command.
[10:40] Right. And so like, what would be the impact if your Grok Bot ran a malicious command?
[10:44] And just to kind of give you like an idea, like if it went to a website and then the website fed it back, you know, malicious content, that could be one access vector.
[10:51] And they listed several. OK, so once the attacker tricks the Grok Bot into running a malicious command, it's basically running inside of like what they call a sandbox.
[11:00] So so step one was, could they break outside the sandbox?
[11:04] And that was trivially easy, right, because they left the there's like a Docker daemon and they let the process in the sandbox talk to the Docker daemon.
[11:15] Right. So, I mean, it sounds ridiculous and it sounds egregious.
[11:20] But I just want to put this in perspective, right, that.
[11:27] When you log into AWS and this is AWS, the world's biggest cloud provider, you can open like a shell in the browser. [Editorial note: AWS CloudShell comparison is host commentary, not independently established by the cited Profero assessment.]
[11:33] Right. And when you open the shell in the browser, it does the exact same thing.
[11:36] It drops you into a sandbox.
[11:37] And when they drop you into that sandbox, you could do the exact same technique and you can drop below the sandbox to kind of the VM that's running below it.
[11:44] So this they're not the only provider out there that is doing this.
[11:48] And the reason that they they do it sometimes, and I don't know if this is the case with Grok Bot, but the case with AWS is sometimes you need to access certain resources or certain secrets.
[11:57] And those are not available in the sandbox.
[11:59] And so the way to do that is to be able to talk to the Docker daemon, which then can like mount file system below it or things like that.
[12:06] So sometimes there is like, OK, there is a reason for it.
[12:09] Now, typically I would recommend building your own custom API that can only do the functions you want it to do.
[12:14] And then don't use the Docker daemon because the Docker daemon has a bunch of other code in it that, you know, you don't want to expose because there might be a vulnerability in that code base.
[12:23] You want to reduce the exposure window as much as possible.
[12:25] But anyways, so I just want to say, like, when I read that, I was like, OK, whatever.
[12:29] I've seen that before with big providers. Right.
[12:31] And this is like a new thing that just came out.
[12:32] So, like, you know, like, I'm not that surprised.
[12:35] But when they got out of the sandbox, they went and looked around to see what secrets they could find.
[12:42] And they found basically you could just use like export the environment variables and you could grab these secrets. [Editorial note: Source nuance: the research reads Docker container configuration from the account VM disk after crossing the sandbox boundary.]
[12:48] And when you could grab these secrets, you could hit the Grok Bot APIs and they would give you back another token.
[12:53] And then you could take that token and talk to Amazon's S3 service, which is what Grok Bot's using for persistent storage.
[13:01] And then you could just download the user files.
[13:03] And the problem with it is that that secret that's outside the sandbox never gets rotated. [Editorial note: Source nuance: long-lived credentials survive sandbox destruction; a VM listener can capture replacement routing tokens. Never rotates/forever is an on-air simplification.]
[13:07] It's always the same. So if someone gets into your Grok Bot environment and just exports those secrets, they can read your files for pretty much forever.
[13:15] Right. So so that's not great.
[13:17] Right. That's that's clearly like a bad design.
[13:20] I mean, it's not like the end of the world. You can definitely change that design, you know.
[13:25] But I got to guess just due to the speed they wanted to get this to market that they cut a few corners.
[13:30] And now those corners are coming to light.
[13:33] So the nice part about this is like there's really nothing that you have to do as a user to fix it. [Editorial note: The source recommends dedicated identities, limited connectors and avoiding production sessions/long-lived keys; no user-action-free mitigation or shipped fix is verified here.]
[13:37] Like, I got to imagine the Grok team is now going to go fix all this stuff. [Editorial note: The expectation of a vendor fix is not evidence one shipped. Profero reported findings remained unfixed as of its November 8 publication.]
[13:41] And they, you know, they recently acquired the cursor company and they're kind of merging those two companies together. [Editorial note: Acquisition/merger and upcoming-version statements are host commentary not verified in this review.]
[13:47] So I also have to guess that they probably have like a V2 of this that they're going to roll out at some point that has some some changes to it.
[13:53] But, yeah, so, you know, I think this is just highlighting, though, just as abstractly the grokness of it.
[14:00] I think it's just highlighting that there's going to be a lot of situations like this where.
[14:05] AI assistants or personal assistants are going to take some specific action that's not going to be secure and really the vendors that are providing these services,
[14:13] or if you're building your own, then you need to be thinking about, like, OK, how do I mitigate and reduce the blast radius zone to be as small as possible?
[14:21] And I do think the vendors will get there over time.
[14:24] But, you know, I do think like if you're doing super mission critical stuff,
[14:28] like you probably don't want to be feeding that data into, you know, experimental services or brand new services that have just come out that have a lot of more features,
[14:37] means more access vectors to exploit and get in and get your data.
[14:43] Yeah, so anyways, I there's a bunch of issues that they highlighted here in the article.
[14:47] If you're interested, we'll link it down below so you can check it out more.
[14:51] But, you know, essentially, this is kind of like.
[14:55] You know, just kind of like.
[14:58] You gave someone the keys to go inside your house and then you like left your safe door open, right?
[15:04] You know, you didn't want them to take things out of the safe, but then you didn't shut the door.
[15:08] So ironically, there was one password that statically set.
[15:11] Do you want to take a guess what the password is to log into this service?
[15:15] Oh, no. Admin.
[15:17] Oh, that's pretty good. But the password is actually local. [Editorial note: In the research, local is the sandbox exec daemon token, not a Grok account login password. Internet access also requires an account-specific routing token.]
[15:20] It's the word local, which, you know, you might not might not be the first.
[15:25] It's not the worst password, but it is definitely up there and about it's definitely up there in the bad password list.
[15:30] So anyways, I.
[15:33] I think the takeaway for here here really is just.
[15:35] If you have like really mission critical data, don't don't be shoving into services that just came out like less than 12 months ago, right? [Editorial note: Age-of-service recommendation and the subsequent 24-month forecast are host opinion/prediction, not established research conclusions.]
[15:41] That's that's my professional recommendation.
[15:45] Because there's going to be more of this.
[15:47] It's going to be like this for the next, you know, 24 months.
[15:50] That would be my prediction.
[15:52] Anyways, Shelby, have you seen anything fun lately that you want to share?
[15:55] Yeah, today I have a game recommendation.
[15:58] I played it yesterday and it was so fun.
[15:59] I forgot how much I like this game.
[16:01] Here it is. It's Magic Maze.
[16:03] And in this scenario, you and your and the other players, it's anywhere from like one to eight players.
[16:09] Nice. But it's funny because you're not like usually your identity is tied to like a piece.
[16:14] Like you're moving the little piece of people around.
[16:16] But your identity is tied to an ability.
[16:19] So each player can like move west.
[16:22] Or north or something like that.
[16:25] And so you can see how you need all the players to be able to navigate your people across the board.
[16:29] And so you're doing a heist and so you can't talk.
[16:32] But there is a timer.
[16:33] And so it's kind of fun.
[16:35] It's good fun. It's it's not hard to learn once you understand the basics of how it works.
[16:40] It's easy to explain to other people.
[16:41] So, yeah, it's really fun.
[16:43] But everyone has to pay attention.
[16:44] No one can zone out because you need all of your abilities going, you know, north, south, east, west and doing a couple other functions.
[16:50] But it's really fun.
[16:51] And you've got to race the clock.
[16:52] And it's very visually stimulating.
[16:55] Like, there's so much to look at on the board because the board changes each game.
[16:58] But, yeah, there you have it.
[16:59] That's my rec.
[17:01] Is it like a collaborative game where you're trying to win together?
[17:03] OK, that's good.
[17:03] So it's a heist and you're like four different mythical creatures who are trying to steal stuff from them all at the same time.
[17:11] Well, I got I got my fun update about a mythical creature, Godzilla.
[17:16] I saw the movie as I was talking about in the last episode that I was going to see it.
[17:19] And I went and saw it. Yeah, it's excellent.
[17:21] I thought they did a really good job.
[17:23] You know, I didn't think I didn't think it was overly silly, like the critics were saying.
[17:26] I thought it was it was good.
[17:29] It was that some good like.
[17:30] It actually has like a plot that makes somewhat sense and while still like having some like roots in the original Godzilla franchise and movies.
[17:38] And, you know, I'm not I'm not the most critical person on movies.
[17:41] So, you know, if you're a critic, then, you know, you may hate it.
[17:44] But I I I enjoyed it. I thought it was good.
[17:47] Bryce Bryce critique. Right.
[17:49] Like there was a lot of plot, right.
[17:53] And there was like some Godzilla blowing stuff up, but not that much.
[17:56] So Bryce critique was I would like it if it was like 90 percent Godzilla blowing stuff up and 10 percent plot.
[18:02] But it was really more like 60 percent plot, 40 percent Godzilla blowing stuff up.
[18:06] So, OK, OK. So I think most people would like this better the way it is.
[18:10] But, you know, I'm also the guy who likes like Fast and the Furious movies and things like that.
[18:14] Like I just like ridiculous stuff.
[18:16] So, you know, I just maybe it's like ADHD type brain, but it's just hard to focus when they're too much.
[18:23] It's been too much plot too long. Right. So, yeah, you're like, move it along.
[18:25] Move it along. Exactly. All right.
[18:27] Well, that's what we got. That's a wrap on today's briefing.
[18:29] A.I. and cybersecurity move super fast and we're here to keep you ahead of it.
[18:33] So stay safe out there and we'll see you next one. Bye.
[18:36] Bye.