The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
No—Paul’s Security Weekly episode 948 does not conclude that AI has taken over cybersecurity. Published on 15 October 2026 after a 14 October recording, the roughly two-hour episode uses AI-agent control, alleged AI worms and incident reporting as a skeptical lens for a much wider security-news roundup. Its practical message is that capability matters less than whether systems can be controlled, patched and contained before technical failures become physical or business crises.
What PSW #948 is—and what it is not
PSW #948 is a current-events podcast episode, not a single-issue AI research paper or a prediction that machines will replace security teams. Paul Asadoorian hosts with Sam Bowne, Josh Marpet, Jeff Mann and Lee Neely. Danny Akacki, known as Rando, is the featured guest.
| Episode detail | Reported information |
|---|---|
| Publication date | 15 October 2026 |
| Recording date | Wednesday, 14 October 2026 |
| Approximate runtime | Two hours |
| Guest | Danny Akacki (Rando) |
| Format | Skeptical security-news roundup with an AI thread |
The episode summary is the basis for the incident descriptions below. The underlying advisories were not independently examined for this article, so figures and event details are identified as claims made in the episode context rather than fresh confirmations.
Why the “AI taking over” question remains open
The AI discussion centers on control and evidence. The hosts examine claims and counterclaims about an Anthropic AI worm, reported limits on Anthropic’s ability to control or monitor AI agents, and the emerging practice of reporting AI-related incidents. Those topics point to a narrower question than “Will AI take over?”: can an agent be constrained, observed and stopped when it behaves unexpectedly?
#1 Best Overall
That distinction separates capability from controllability. A system that can write code or operate tools may still be manageable if permissions, logging, isolation and human review work. Conversely, a modest capability can cause serious damage when it is connected to production systems without segmentation or a reliable shutdown path. Josh Marpet’s line captures the episode’s emphasis: “It’s the consequences that is the problem.”
The show also treats sensational claims cautiously. An alleged AI worm and an AI incident report require evidence about what happened, which systems were reached, what autonomy was actually demonstrated and what safeguards failed. The episode does not present either claim as proof of an autonomous takeover.
The security stories, grouped by the risk they expose
Internet exposure, patching and root-level compromise
Several items connect a known technical weakness to an avoidable exposure window:
Rank #2
- A router was converted into a software-defined radio, illustrating how ordinary network equipment can be repurposed when administrators overlook hardware capability and firmware controls.
- An Ivanti issue was described as a straight-to-root compromise, making the difference between a limited foothold and complete device control immediate.
- A kernel exploit was reportedly used to fix downloads, an example of offensive techniques being turned toward repair or recovery rather than only intrusion.
- The episode discussed more than 400 Java vulnerabilities alongside IBM and Red Hat’s Lightwell Clearinghouse, a paid dependency-review service intended to help organizations identify risky components.
- FortiBleed coverage included a figure of 86,000 firewalls. That number is attributed to the episode’s campaign discussion and to Podcast Rex, 2026; it should be checked against the original incident reporting before being used as a standalone statistic.
- A critical GitLab patch reinforced the same operational rule: disclosure is only the beginning; affected installations still need inventory, testing and deployment.
Paul Asadoorian’s blunt reaction—“I could have wrote the shell script to do this while I was taking a shit”—underscores how routine some remediation work can be once the vulnerable systems are known, even when the consequences of delay are not routine.
Operational technology and physical consequences
The refrigeration story is the clearest warning that cybersecurity is not confined to data. French supermarket refrigeration units reportedly reached 122 degrees Fahrenheit after internet exposure and a lack of patching. The 122-degree figure is an episode-summary claim attributed to Podcast Rex, 2026, not an independently verified measurement here. The consequence is nevertheless easy to understand: an exposed control system can spoil food and disrupt stores even when no customer database is stolen.
Other operational-technology stories broaden that lesson:
- An authentication bypass in Sungrow solar infrastructure shows how an identity failure can affect energy equipment.
- A Russia-related data-center conflict links geopolitical pressure to availability and physical infrastructure.
- An incident involving Stop & Shop and parent company Ahold Delhaize shows how retail operations can be disrupted at scale.
- The hosts discussed an engineer allegedly locking thousands of devices, a reminder that privileged insiders and administrative mistakes can have a blast radius comparable to an external attacker.
Lee Neely’s dry sign-off—“I hope you’ve enjoyed the demo. Please stand by for full deployment”—fits this category: a proof of concept becomes a crisis when it reaches systems that run stores, energy assets or customer operations.
Identity, deception and software trust
The roundup also covers attacks and defenses that depend on believing the wrong thing:
- Pre-compromised phones raise the supply-chain question of whether a device is trustworthy before an owner signs in.
- FakeGit targets the trust developers place in repositories, names and familiar workflows.
- Teams deepfake detection addresses synthetic voices or video that can make a fraudulent request appear to come from a colleague.
- Let’s Encrypt’s move toward shorter certificate lifetimes means automation and renewal monitoring matter more than treating certificates as set-and-forget assets.
- The episode includes an allegation involving an OpenAI employee and a hidden recorder. It is presented as an allegation, not an established finding.
Together, these items favor verification over appearance: verify the source of code, the identity behind a message, the provenance of a device and the current validity of a certificate.
Danny Akacki and DEF CON 610
Akacki is described as a DEF CON goon and a client security strategy director in Coalfire’s Hex division. He introduces himself with “Long time listener, first time crasher.” In another humorous moment he says, “I didn’t know that Huawei was a legitimate company. I thought they were a threat actor group,” reflecting the show’s conversational style rather than a technical conclusion.
He also discusses DEF CON 610’s Net Run CTF, scheduled for 21 November 2026 in Easton, Pennsylvania. The event replaces the former pub-crawl format. Anyone planning around it should confirm the organizer’s current schedule and participation requirements, since the date is the episode’s listed schedule.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteProducts, services and reading named in the episode
The conversation mentions several concrete resources, but naming does not establish an endorsement, current price or availability:
Best Value
- JetKVM: a physical KVM product and the episode’s strongest hardware-related fit.
- Gemtek W1700K: another specific hardware query raised in the product discussion.
- More Everything Forever by Adam Becker: a named book recommendation.
- ThreatLocker: appears in a sponsor read.
- Lightwell Clearinghouse: described as a paid dependency-review service associated with IBM and Red Hat.
Marketplace listings, sponsorship terms, affiliate status and commissions were not established in the episode material, so readers should verify those details directly before buying or subscribing.
What a security team can take from the episode
- Map internet-facing assets. Include routers, solar controllers, refrigeration and retail systems, not just servers and laptops.
- Prioritize fixes by consequence. A root-level appliance flaw, an authentication bypass or an unpatched control system deserves faster action when failure can stop operations or damage goods.
- Segment operational technology. Internet exposure should not provide a direct path to equipment that controls temperature, power or store operations.
- Make dependency and certificate upkeep automatic. Review third-party Java components and build renewal monitoring for shorter certificate lifetimes into deployment pipelines.
- Challenge AI claims with incident evidence. Record the agent’s permissions, tools, actions, human checkpoints and demonstrable impact before labeling behavior an autonomous worm or takeover.
- Harden identity and provenance checks. Treat repository names, pre-owned devices, collaboration messages and synthetic media as sources that require verification.
The episode’s bottom line
PSW #948’s answer to “Is AI taking over cybersecurity?” is a qualified no: the urgent issue is not a dramatic transfer of control to AI, but whether people can govern increasingly capable tools while ordinary weaknesses remain exposed. The episode’s router, refrigeration, solar, retail and dependency stories all point to the same test—control the system, patch it promptly and measure the real-world consequence when something goes wrong.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
