Earlier this week, OpenAI dropped GPT-5.4-Cyber, a “cyber-permissive” variant of GPT-5.4 .&nbsp;&nbsp;&nbsp; Basically: It's lets you do some things in the name of security research and defense that you can’t normally do with a regular GPT model.&nbsp;&nbsp; But you have to prove you’re a cybersecurity pro with good intentions to get access.&nbsp; On this bonus episode of Security Intelligence, Jeff Crume and Martin Keen join host Matthew Kosinski to break down what the new model means for cybersecurity and the big picture trends driving the evolution of LLMs.&nbsp;&nbsp; The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.&nbsp; #GPT-5.4Cyber #OpenAI #AISecurity Follow the Security Intelligence podcast on your preferred platform → <a href="https://www.ibm.com/think/podcasts/security-intelligence" rel="noopener noreferrer" target="_blank">https://www.ibm.com/think/podcasts/security-intelligence</a>&nbsp;

OpenAI’s GPT-5.4-Cyber is a new “cyber-permissive” variant model designed for cybersecurity. But who gets to use it–and how? 

GPT-5.4-Cyber: What you need to know

We’ve&nbsp;all heard tales of what lurks on the dark web. Black markets dealing in contraband. Roving criminal gangs. Troves and troves of&nbsp;private data&nbsp;for sale. Much, much worse.&nbsp; Thankfully, most of&nbsp;it’s&nbsp;fake. The real trouble is figuring out what&nbsp;isn’t.&nbsp; In this episode of&nbsp;Security Intelligence, Senior Threat Intelligence Analyst Robert Gates helps us untangle the web of lies that cybercriminals spin on the dark web—the lies they tell their victims, their customers, and even themselves.&nbsp; By walking us through a few real-world investigations—including a panicked airline that thought it had been breached and the still-unsolved case of a hacker who may or may not have stockpiled millions of sensitive records—Robert shows us how threat intelligence experts and cybersecurity pros separate fact from fiction.&nbsp; The dark web can be your most reliable early warning system. But only if you know how to use it.&nbsp; The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.&nbsp; Read more about the dark web → <a href="https://www.ibm.com/think/topics/dark-web" rel="noopener noreferrer" target="_blank">https://www.ibm.com/think/topics/dark-web</a>&nbsp;Follow the Security Intelligence podcast on your preferred platform&nbsp;→&nbsp;&nbsp;<a href="https://www.ibm.com/think/podcasts/security-intelligence" rel="noopener noreferrer" target="_blank">https://www.ibm.com/think/podcasts/security-intelligence</a>&nbsp;

Most of what you’ve heard about the dark web is fake. The real trouble is figuring out what’s true.  

Web of lies: What’s real and what’s fake on the dark web

Sophos let&nbsp;OpenClaw&nbsp;run wild on its network (sort of). It&nbsp;wasn’t&nbsp;as bad an idea as it sounds!&nbsp; With a few guardrails and restrictions in place, the security software firm turned&nbsp;OpenClaw&nbsp;into a serious little pen tester, surfacing “23 actionable, high-quality findings.”&nbsp; But is this a sustainable model for introducing AI agents to the security process? And how do we deal with the inevitable friction between a model meant to find exploits and the guardrails telling it to do no harm?&nbsp; This week, host Matt Kosinski and panelists Claire Nuñez, Dave McGinnis and Kimmie Farrington discuss the wisdom and folly of letting an AI agent pen test your system.&nbsp; Plus: We dig into Bruce Schneier’s thoughts on “security in the age of instant software” and a report from&nbsp;CipherCue&nbsp;that ransomware is growing three times faster than security spending.&nbsp;&nbsp; All that and&nbsp;more&nbsp;on&nbsp;Security Intelligence.&nbsp; Segments:&nbsp; 00:00 – Intro&nbsp; 1:07 --&nbsp;OpenClaw&nbsp;as a pen tester&nbsp;&nbsp; 14:23 -- Cybersecurity for instant software&nbsp;&nbsp;&nbsp;25:36 -- Ransomware outpaces security spending&nbsp;&nbsp;The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.&nbsp;&nbsp;Learn more about how enterprises confront agentic attacks&nbsp;→&nbsp;<a href="https://newsroom.ibm.com/2026-04-15-ibm-announces-new-cybersecurity-measures-to-help-enterprises-confront-agentic-attacks" rel="noopener noreferrer" target="_blank">https://newsroom.ibm.com/2026-04-15-ibm-announces-new-cybersecurity-measures-to-help-enterprises-confront-agentic-attacks</a>&nbsp;Follow the Security Intelligence podcast on your preferred platform&nbsp; →&nbsp;<a href="https://www.ibm.com/think/podcasts/security-intelligence" rel="noopener noreferrer" target="_blank">https://www.ibm.com/think/podcasts/security-intelligence</a>&nbsp;

AI agents as pen testers, ephemeral software risks and ransomware outpacing security budgets 3-to-1. IBM's Security Intelligence breaks it down. 

Should you let OpenClaw pen test your system? Plus: Cybersecurity for ephemeral software

Where does that leave the rest of us, who don’t get to tinker with perhaps the most advanced&nbsp;model yet?&nbsp; This week on Security Intelligence, Sridhar Muppidi, Michelle Alvarez, and Dustin “EvilMog” Heywood join host Matt Kosinski to discuss what Mythos and Glasswing really mean for the average security pro. &nbsp; How much is hype? How much is the real deal? And how could this limited release backfire?&nbsp; Then: The FBI’s 2025 Internet Crime Report saw scam losses jump 26%, and Accenture found a 127% increase in malicious hackers trying to recruit the employees of their targets. &nbsp; All that and more on Security Intelligence.&nbsp; 00:00 – Intro&nbsp;1:22 -- Claude Mythos and Project Glasswing&nbsp;12:26 -- The 2025 Internet Crime Report&nbsp;20:19 -- Attackers recruiting more insiders&nbsp; The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.&nbsp; Read more about AI raising cybersecurity stakes → <a href="https://www.ibm.com/think/news/anthropic-claude-ai-mythos-project-glasswing-raises-stakes-cybersecurity" rel="noopener noreferrer" target="_blank">https://www.ibm.com/think/news/anthropic-claude-ai-mythos-project-glasswing-raises-stakes-cybersecurity</a> &nbsp;Follow the Security Intelligence podcast on your preferred platform → &nbsp;<a href="https://www.ibm.com/think/podcasts/security-intelligence" rel="noopener noreferrer" target="_blank">https://www.ibm.com/think/podcasts/security-intelligence</a>&nbsp;

Claude Mythos is so good at finding software vulnerabilities that Anthropic will only release it to a handful of partners through Project Glasswing. What about the rest of us? 

Claude Mythos: Marketing hype or the end of cybersecurity?

What happens when one of the world’s most popular AI coding tools falls into the wrong hands?&nbsp; On this episode of Security Intelligence, Nick Bradley, Dave Bales and JR Rao discuss the Claude Code source code leak. Attackers are already using the opportunity to spread malware through fake repos, but the real question is how threat actors might use their newfound knowledge of Claude Code’s internals to wreak havoc on AI agents and the CI/CD pipeline.&nbsp; Then, we follow up on our old friends TeamPCP, Shiny Hunters and Lapsus$, whose overlapping data breach claims are causing no small amount of confusion and consternation among security pros. We examine the credential rotation problem and the uneven security surface of modern supply chains that helped get us in this mess.&nbsp; Plus: Threat intelligence usually focuses on attacks that did happen. But what if we started talking about the ones that didn’t? And do cybercriminals have anything to teach us about “mature” AI adoption? Some big names seem to think so.&nbsp; All that and more on Security Intelligence.&nbsp; Segments:&nbsp; 00:00 – Introduction1:12 -- The Claude Code leak&nbsp;11:19 -- TeamPCP’s breach spree&nbsp;21:21 -- “Close-call” databases&nbsp;&nbsp;29:28 -- Cybercrime and AI adoption&nbsp; The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.&nbsp; Visit the Security Intelligence the podcast page → <a href="https://www.ibm.com/think/podcasts/security-intelligence" rel="noopener noreferrer" target="_blank">https://www.ibm.com/think/podcasts/security-intelligence</a> Explore to securely deploy and operate agentic AI workloads at runtime → <a href="https://ibm.webcasts.com/starthere.jsp?ei=1755597&amp;tp_key=10f0b8919a&amp;sti=inbound" rel="noopener noreferrer" target="_blank">https://ibm.webcasts.com/starthere.jsp?ei=1755597&amp;tp_key=10f0b8919a&amp;sti=inbound</a>

The Claude Code source code leak: Takeaways for cybersecurity pros

﻿Learn more about solving agentic AI identity and access gaps → <a href="https://www.hashicorp.com/en/blog/agentic-runtime-security-solving-agentic-ai-identity-and-access-gaps" rel="noopener noreferrer" target="_blank">https://www.hashicorp.com/en/blog/agentic-runtime-security-solving-agentic-ai-identity-and-access-gaps</a> LiteLLM is a nifty little Python library that gives you access to about 100 different AI services through one API. It gets an estimated 3.4 million downloads a day.&nbsp; And last week, it was turned into a Trojan horse, distributing infostealers to hundreds of thousands of devices. (At least, that’s what TeamPCP says—the hackers behind the LiteLLM breach and a slew of other high-profile software supply chain attacks in recent weeks.)&nbsp; Quote Andrej Karpathy: This is “basically the scariest thing imaginable in modern software.”&nbsp; On this episode of Security Intelligence, Suja Viswesan, Dave McGinnis and Jeff Crume help us break down the LiteLLM breach and the broader campaign TeamPCP is waging.&nbsp; We’re also joined by HashiCorp Field CTO Jake Lundberg in the first segment for a discussion of how organizations are trying—with varying degrees of success—to tackle the agentic AI problem.&nbsp;&nbsp; AI agents are identities—but identities our existing frameworks weren’t built to house. Simply porting existing human and non-human identity management practices onto them won’t cut it.&nbsp; But the question remains: What do we need instead?&nbsp; All that and more on Security Intelligence.&nbsp; Segments&nbsp; 00:00 -- Intro&nbsp;1:13 -- Who will fix AI agent security?&nbsp;&nbsp;21:17 -- RSAC 2026 Recap&nbsp;&nbsp;29:31 -- 2026's most dangerous cyberattacks&nbsp;&nbsp;40:45 -- The LiteLLM breach&nbsp;&nbsp; The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.&nbsp; Explore the podcast → <a href="https://www.ibm.com/think/podcasts/security-intelligence" rel="noopener noreferrer" target="_blank">https://www.ibm.com/think/podcasts/security-intelligence</a>

The latest in AI agent security, the LiteLLM breach, key takeaways from RSAC 2026 and SANS’s most dangerous cyberattacks of the year. 

RSA recap, the LiteLLM breach, and the quest to fix AI agent security

Most cybersecurity pros only run into cryptocurrency when they’re dealing with ransomware gangs demanding payouts in Bitcoin.&nbsp;&nbsp; But what if crypto infrastructure were more than just a means of money laundering?&nbsp; In this episode of IBM’s Security Intelligence podcast, X-Force threat intelligence consultant Austin Zeizel makes the case that blockchain — the decentralized ledger underlying many cryptocurrency systems — has powerful, largely untapped applications for cybersecurity. In fact, Austin makes a pretty convincing argument that blockchain could be the key to a genuinely zero trust architecture.&nbsp; We also get into how cybercriminals actually exploit cryptocurrency — coin mixers, non-KYC exchanges, the pseudonymous nature of Bitcoin — and why understanding those mechanics matters for defenders.&nbsp; The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.&nbsp;

Cryptocurrency isn’t just ransomware gangs’ favorite payment. It might also be the key to finally realizing genuine zero trust. 

Cryptocurrency: The most misunderstood technology in cybersecurity

Security Intelligence is a weekly news podcast for cybersecurity pros who need to stay ahead of fast-moving threats. Each week, we cover the latest threats, trend, and stories shaping the digital landscape, alongside expert insights that help make sense of it all. Whether you’re a builder, defender, business leader or simply curious about how to stay secure in a connected world, you’ll find timely updates and timeless principles in an accessible, engaging format. New episodes weekly on Wednesdays at 6am EST.

IBM

GPT-5.4-Cyber: What you need to know

DESCRIPTION

Today's Host

Matt Kosinski

Recent Episodes