Steve Gibson

Things Steve Says on Podcasts

Software developer and security researcher who founded Gibson Research Corporation, maker of the SpinRite disk maintenance and recovery utility and the ShieldsUP! internet security testing service. Co-hosts Security Now with Leo Laporte on the TWiT network, a weekly show on security news, patch cycles and the risks of AI agents.

Where to Find Them

Steve Gibson hosts Security Now , hosts Security Now (Audio) , hosts All TWiT.tv Shows (Audio) , hosts Security Now (Video) , hosts All TWiT.tv Shows (Video) , hosts TWiT Events (Audio) , hosts TWiT Throwback (Video) and hosts TWiT Events (Video) . They have also been a guest on Padre's Corner (Audio) .

Recently: “SN 1105: Formal Code Verification - When All Your Apps Go Dark” on Security Now (Audio) (November 2026); “SN 1105: Formal Code Verification - When All Your Apps Go Dark” on Security Now (November 2026); “SN 1105: Formal Code Verification - When All Your Apps Go Dark” on Security Now (Video) (November 2026); “Security Now 1105: Formal Code Verification” on All TWiT.tv Shows (Video) (November 2026); “Security Now 1105: Formal Code Verification” on All TWiT.tv Shows (Audio) (November 2026); “SN 1104: The hole in OpenSSH - Password Bypass in Plain Sight” on Security Now (November 2026).

What They Said

“So much of our infrastructure has been put on the internet under the fragile assumption that the device's authentication front end is flawless. Unfortunately, now more than ever, flawless is what it needs to be. Yet we keep encountering example after example of readily bypassable mistakes in authentication. Authentication is hard. Apparently, it's impossibly hard.” — Steve Gibson, Security Now

This closes Steve Gibson's coverage of attacks on internet-connected industrial refrigerators in France, and leads into the episode's main story: a long-hidden authentication bypass in OpenSSH on musl-based Linux.

Security Now · 2026-11-11 Permalink → Listen →
Security Now Around 1:26:48 into the episode
Steve Gibson

And, you know, it's like, now you do. Well,

Leo Laporte

you had journal control. You had syslog. You know, you were logging everything that happened on the machine, but that's not enough with these AI agents. They're little guys running around inside there and they're doing stuff. And you need to know what they're doing. Absolutely. On we go.

Steve Gibson

…systems, including systems in France. So this is the sort of story with the, as I said before, with the wide coverage that it received that will, let's hope, work to get other internet-connected device makers to re-examine their security or take their security more seriously or buy some compute from OpenAI. Fuck it if you have to to check the security of your devices before the bad guys do. So much of our infrastructure has been put on the internet under the fragile assumption that the device's authentication front end is flawless. Unfortunately, now more than ever, flawless is what it needs to be. Yet we keep encountering example after example of readily bypassable mistakes in authentication. Authentication is hard. Apparently, it's impossibly hard. And in fact, that's what today's main topic will be about. And as I said, it is a doozy, which we will get to shortly. First, Leo, we're going to take a break and then look at feedback from our listeners.

Leo Laporte

Indeed. All right. Well, thank you, Steve. Our show today brought to you by Delete Me. We actually use DeleteMe to protect our management from hacker emails. Your business invests in cybersecurity, I'm sure. That's why you listen to the show. But what about the personal information cyber criminals…

Steve Gibson

Okay, so Don Edwards writes, Dear Steve Gibson must take issue with the notion you discussed last week, episode 1103, about a web server behind the protection of the Cloudflare WAF, the web application firewall, being, quote, off the internet, unquote. Much like AV software, the WAF can only offer…

Leo Laporte

it to get out there. Wow. What

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“I've always modeled security as an inherently porous medium, the idea being that the harder you push, the more vulnerable a system would be shown to be. But AI is revealing that even our most hardened software is softer than we knew.” — Steve Gibson, Security Now

Steve Gibson was going through an AI lab's cyber-model announcement that included dozens of new high-severity Firefox bugs confirmed by Mozilla, in code he calls among the most scrutinized in the world. He had expected AI to run out of flaws in well-reviewed software faster than it is.

Security Now · 2026-11-11 Permalink → Listen →
Security Now Around 57:26 into the episode
Steve Gibson

So it's a good question.

Leo Laporte

Maybe you better talk, start with talking about what that is.

Steve Gibson

Yeah, not as I understand it. So as I noted last week, I have, and I mentioned this last week, I've been very much enjoying Kevin Roos's New York Times best-selling book, The AGI Chronicles. It sounds like you were familiar with it, Leo. Oh, yeah, absolutely. Okay. So I'm still only 40% of the way through the book, but I've noticed that now, when I'm reading the news about the various people and personalities who are making AI news, I know so much more about them than, you know, just their name. I have a real sense for who they are, which is fun. And that's the reason I picked this book up is I wanted to understand the backstory. This all to us who weren't paying attention to what was going on for the last decade, it feels like this suddenly just AI just sprang. But there's a lot more that happened historically. One of the terms that I learned from the book is the phrase AI pilled. It's a derogatory term used within the AI community. It's sort of meant in the sense of the Matrix's Neo, right? Who was offered either…

Leo Laporte

One of the things maybe it's Marcus Hutchins was saying this on X that most hackers are not sophisticated enough to really be able to use these AI tools in the way that we worry about. That this is still not script kitty material level. Right. Right.

Steve Gibson

Yeah.

Leo Laporte

Right. And I think that that's encouraging. And they're not, and we've seen that. The attacks, we've seen a lot of flaws that we found and fixed, but the attacks are not swamping us by any means yet.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“I don't think I'm becoming AI-pilled, but I am increasingly becoming AI-stunned and AI-awed. … The more I push the limits of what Claude can do, the fuzzier the presence of any limit feels. I haven't found a limit yet. I'm just astonished.” — Steve Gibson, Security Now

Steve Gibson had just explained that 'AI-pilled' is a derogatory term for someone who believes AI is about to become sentient. After days of building with Claude, he denies being AI-pilled but is clearly impressed; Leo Laporte answered that he is AI-pilled for sure.

Security Now · 2026-11-11 Permalink → Listen →
Security Now Around 57:26 into the episode
Steve Gibson

So it's a good question.

Leo Laporte

Maybe you better talk, start with talking about what that is.

Steve Gibson

…reports that in-house AI researchers are always experimenting with something next, sometimes way next. You know, they have significantly more advanced models than have yet been widely released for public use. So, my point here is that the more I push the limits of what Cloud 5.5 can do, the fuzzier the presence of any limit feels. I don't, I haven't found a limit yet. I'm just astonished. So, I don't think I'm becoming AI-pilled, but I am increasingly becoming AI-stunned and AI-awed. And I'm certain that anyone who has used the latest AI extensively, and I know you have, Leo, like more than extensively, you'll understand what I'm saying. I mean, it is astonishing what we have. You know, and when I, you know, the industry is getting a little wobbly at the moment. You know, we have all this Jensen nonsense going on, but, you know, and a concern about profit, a concern about…

Leo Laporte

One of the things maybe it's Marcus Hutchins was saying this on X that most hackers are not sophisticated enough to really be able to use these AI tools in the way that we worry about. That this is still not script kitty material level. Right. Right.

Steve Gibson

Yeah.

Leo Laporte

Right. And I think that that's encouraging. And they're not, and we've seen that. The attacks, we've seen a lot of flaws that we found and fixed, but the attacks are not swamping us by any means yet.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“We're now going to see cryptocurrency insurance rates going up as insurers are being forced to make big payouts. It's the same thing we saw with ransomware insurance: their rates went way up, so that it became very difficult to afford.” — Steve Gibson, Security Now

After the $6.3 billion Binance theft, co-host Mikah Sargent wondered how many thefts it will take before the crypto space gets better protections. Gibson predicts that insurance pricing will be the pressure that forces change, as it did with ransomware.

Security Now · 2026-11-04 Permalink → Listen →
Security Now Around 1:01:38 into the episode
Steve Gibson

…wallets when they lost bets. It does make sense that some sort of subversion of the Polymarket platform could transfer away their users' walleted funds. So it looks like some guy, someone somewhere, some party got $2.7 million and this platform is going to be covering it out of their own assets.

Mikah Sargent

I wonder how many it's going to take before we see some more, I don't know, protections in place in the crypto space.

Steve Gibson

And probably what we're going to begin to see, and it's the same thing that we saw in cybersecurity insurers who were being forced to pay out a lot of money when ransomware was extorting money from their clients, their insurance rates went way up so that it became very difficult to afford cybersecurity, you know, ransomware insurance. We're now going to see cryptocurrency insurance rates going up as insurers are being forced to make big payouts. That'll happen. You know, the other thing that is going to happen right now, Micah?

Mikah Sargent

Will be an ad break. Yeah. Let's take a break and hear from Leo.

Speaker 4

All right. This episode of Security Now brought to you by Trusted Tech. I hope you're listening because if you're using Microsoft Unified Support, there's something you ought to know. It's priced as a percentage of what you spend with Microsoft. Not what you need, but what you spend with Microsoft…

Mikah Sargent

All right. Thank you, Leo. Let's continue with security now.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“This means that my saying, as I did last week, that GRC would be a low-interest target really doesn't fly, because the threshold required for interest has been lowered to the floor, so that no one is low interest any longer.” — Steve Gibson, Security Now

Gibson had just read a listener's account of his 30-person startup being hacked, apparently by an internet-wide automated scan with AI digging into the services it found. He uses it to revise his earlier claim that his own company would not be worth attacking.

Security Now · 2026-11-04 Permalink → Listen →
Security Now Around 1:30:38 into the episode
Steve Gibson

…knowledge that we have into language, which we have long recorded, right? On cave walls, stone tablets, parchment, and now in books, and actually now in LLMs. So, what happened during the past few years is that we finally figured out how to encode and record that same knowledge in neural networks.

Mikah Sargent

…comes from us? Or is it enough that our observation of said, because otherwise, how could the author that wrote their book, how would they be able to have that knowledge if it wasn't from getting it through information elsewhere? And if that's the case, then upon observation, it becomes knowledge.

Steve Gibson

And that's my point. I mean, I can see Guillermo's, I can see the point he wants to make and the position he takes. For me, I don't think I will change my way of thinking about it or expressing it. You know, a book, if I say that a book contains knowledge, to me, that feels right because that's what I obtain. And I mean, entertainment and humor and emotion. I mean, it contains a lot of stuff, right? Yes. It can impart knowledge. So I think it contains that knowledge that it's imparting. And it contains the humor that's imparting and the emotional affect that it is imparting. So, you know, we're affected by those things because they're in there. Right. Yeah, and they are expressed in and encoded into language. But that's knowledge that is in there that is in the book, I think. So Jordan B. So Jordan used his full name and did not ask me to obscure his identity, but it kind of felt like the right thing to do. Just because for propriety's sake, he gave his email, which I received last Tuesday morning, the…

Mikah Sargent

made that it's good that it happened while they were a small business and not a large business. And better now than later. Yeah. Yeah, when there was much

Steve Gibson

more, much more to lose.

Mikah Sargent

Exactly. And be fined for.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“The genie has not only escaped the lantern, the lanterns have been mass produced and everybody has their own genie. So why bother?” — Steve Gibson, Security Now

A listener asked what a functional kill switch for AI would look like. Gibson argues no universal switch is possible because people already run capable AI locally and in personal clouds, so shutting down the commercial providers would soon be pointless.

Security Now · 2026-11-04 Permalink → Listen →
Security Now Around 1:37:53 into the episode
Steve Gibson

more, much more to lose.

Mikah Sargent

Exactly. And be fined for.

Steve Gibson

…two years from now in 2028 with our next president. Given the incredible rate of progress that we've seen from AI so far, two years. My God, that's a lifetime. And by that point, I suspect that so much ultra-capable AI will be available and running locally and also in personal cloud infrastructure that cutting off the commercial providers will be seen by then as pointless because it's like the genie has not only escaped the lantern, the lanterns have been mass produced and everybody has their own genie. So why bother?

Mikah Sargent

That's a good way to put it.

Steve Gibson

And suddenly, or finally, Asher says, hello, Steve. I've heard people saying they want more AI stuff on the show and some want less AI on the show. For me, it's more on the type of AI coverage. At this point, the weekly update on what amazing new trick the AI has learned is no longer interesting.…

Mikah Sargent

All right. Let us take a moment here to pause and I will just remind you all about Club Twitch. At twit.tv slash club twit, which is where you can go to become a member of the club. When you do so, you gain access to some pretty awesome benefits. Every single one of our shows, ad-free, just the…

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“My own theory, based upon watching the use of vulnerability exploits for many years, is that disrupting or taking down the internet is not profitable. What is highly profitable is breaking into an organization, exfiltrating their data, and then extorting payment from the breached organization in return for that data's deletion.” — Steve Gibson, Security Now

Gibson was considering what attackers with malicious intent will do with AI. He points out that flaws capable of producing an internet-wide worm have existed for years without anyone using them that way. His conclusion is that AI will speed up the attacks that already pay, and that wrecking the internet would cut off the attackers' own income.

Security Now · 2026-09-30 Permalink → Listen →
Security Now Around 2:21:01 into the episode
Steve Gibson

Your computer reached out and made connections. Mistakes

Leo Laporte

…cutting edge of AI on Sun Friday, 2 p.m. Pacific. That's an example of what the club does. If you want to help us out, twit.tv slash club twit. We sure would appreciate your support. If you like what Steve's doing, that's the best way to support it. All right. On we go with the show, Mrs. So as an

Steve Gibson

…exposed Windows server products that could have been used to create a devastating internet flashworm. But that didn't happen, nor did it happen. Last month or the month before, and I doubt it will happen next month. For some time, we've seen serious vulnerabilities publicly exposed in Cisco Edge border routers that could have been leveraged to do the same thing, but that's never occurred. My own theory, based upon watching the use of vulnerability exploits for many years, is that disrupting or taking down the internet is not profitable. What is highly profitable is breaking into an organization, exfiltrating their data, and then extorting payment from the breached organization in return for that data's deletion. So, that is what has been going on. And there's every reason to believe that's what will continue to go on. Today's AI will likely serve as nothing more than an accelerant poured over the present status quo. I don't think it's going to see the world change. So, I just expect that we're going to be seeing more of the same from the malicious use of AI rather than anything apocalyptic. And once the…

Speaker 4

Nothing to worry here. Move on. You know, I think the question for you, Doug, would be: do you

Leo Laporte

acknowledge that there are benefits to be achieved from this technology? And I think that that's the thing Steve and I both are saying is that we can see significant, I already see significant benefits, and I see many, many more coming down the road. If you see benefits to this, yes, there's also…

Steve Gibson

Who let those things off their tracks, Leo?

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

Collections They Appear In