Cybersecurity Today: New Threats from AI and Code Extensions

Summary5 min read

Host: Jim Love
Date: October 24, 2025

Episode Overview

This episode of Cybersecurity Today dives into the rapidly evolving threat landscape, with a particular focus on how AI-powered attacks and supply chain compromises via code extensions are outpacing traditional defenses. Jim Love breaks down striking new threats — including the self-propagating "Glass Worm" malware in Visual Studio Code extensions, widespread vulnerabilities in AI-powered IDEs, and the rising fear of AI-driven attacks as captured in the latest ISACA survey. The conversation is rich with actionable advice and sobering analysis for cybersecurity professionals and organizations wanting to safeguard their operations in 2026.

Key Discussion Points and Insights

1. Glass Worm: A Sophisticated Supply Chain Attack

[01:30 – 06:20]

Discovery and Spread:
- Glass Worm, a self-spreading malware, was hidden in Visual Studio Code (VS Code) extensions.
  - Over 35,000 downloads before detection.
  - Discovered by researchers at Koi Security; found in seven extensions on Open VSX Marketplace and at least one in Microsoft's own VS Code Marketplace.
- Quote [02:11]:
  
  "What makes Glassworm so dangerous is how well it hides — the malicious JavaScript is written using invisible Unicode characters, variation selectors that show up as blank space in the editor. To a developer it looks like empty lines, to scanners it looks like nothing, but to AI, it’s executable code."
  — Jim Love
Capabilities and Impacts:
- Steals developer credentials (GitHub, Git, NPM).
- Drains funds from crypto wallets.
- Converts compromised machines into proxy servers.
- Installs covert remote access tools.
- Employs Solana blockchain and Google Calendar for command and control, making takedowns difficult.
Immediate Response Recommendations:
- Security teams using VS Code should:
  - Treat the situation as an active incident.
  - Inventory and review all installed extensions against the list posted by Koi Security.
  - Disable automatic updates and scrutinize every extension.
  - Restrict access to untrusted extension marketplaces and whitelist only approved extensions.
Memorable Moment [05:00]:

"For the moment, the safest move is to block untrusted marketplaces and allow list only approved extensions until the threat is fully contained."
— Jim Love

2. Vulnerabilities in AI-Powered IDEs: Cursor and Windsurf

[06:21 – 09:35]

Critical Risks Identified:
- Newly-released AI-powered IDEs, Cursor and Windsurf, are based on outdated Chromium and Google V8 engines.
- More than 94 security flaws, including critical ones, affect the embedded browsers.
- Over 1.8 million developers potentially at risk.
Exploitation Example:
- The integer overflow flaw, CVE-2025-7656, could crash the editor and potentially allow remote code execution via a malicious link.
Vendor Responses (or Lack Thereof):
- Cursor labeled the exploited vulnerability “out of scope."
- Windsurf did not respond at all.
Quote [08:22]:

"Electron-based apps like these package a fixed Chromium build. Unless developers regularly update that embedded browser, they inherit every browser vulnerability that's been fixed since their last release. In this case, that means 94 of them, some of them critical."
— Jim Love
Recommendations:
- Organizations using these IDEs should:
  - Prioritize this as an incident-level concern.
  - Pressure vendors for patched, up-to-date builds.
  - Consider alternative tools if vendors dismiss or ignore these issues.

3. AI-Driven Attacks Become Top Security Fear for 2026

[09:36 – 13:05]

ISACA 2026 Tech Trends and Priority Pulse Poll:
- AI-driven threats have overtaken ransomware and insider breaches as top concerns.
- Statistics:
  - 59% cite AI and deepfakes as their primary worry.
  - 63% highlight AI-driven social engineering (cloned voices, fake videos, automated spear-phishing).
Current Level of Preparedness:
- Only 13% of organizations feel “very prepared.”
- Nearly one-third admit they’re not ready at all.
Quote [11:23]:

"Attackers are already using AI to craft convincing messages, mimic executives, and scale deception in ways human criminals never could. Only 13% of respondents said their organizations are very prepared to handle these risks."
— Jim Love
Pace of Change:
- 41% of professionals report that just keeping up with AI's pace is their biggest challenge.
- There is a growing mismatch between the speed of AI adoption and the ability of security teams to build effective defences.
Quote [12:16]:

"The technology that’s transforming our business is now transforming the threat landscape faster than most organizations can respond."
— Jim Love

Notable Quotes & Memorable Moments

On New Forms of Malware Stealth:

"...to scanners it looks like nothing, but to AI, it’s executable code." [02:20]
On Supply Chain Security Recommendations:

"...the safest move is to block untrusted marketplaces and allow list only approved extensions until the threat is fully contained." [05:00]
On the Failure to Patch IDE Security Flaws:

"If you hear that that's out of scope or you don't get an answer, maybe it’s time to go looking for a new tool set." [09:10]
On the Strain Facing Security Professionals:

"The speed of AI adoption is outpacing the defenses meant to contain it, and the strain is starting to show..." [11:55]

Timestamps for Important Segments

[01:30] Unveiling of Glass Worm in VS Code extensions
[03:00] Glass Worm’s capabilities & command and control mechanisms
[04:45] Guidance for organizations to assess and secure code extensions
[06:21] Widespread vulnerabilities in AI-based IDEs exposed
[07:50] Vendor responses and the inherent risks of Electron-based applications
[09:36] ISACA survey: AI-driven threats lead 2026 concerns
[11:20] Organizational preparedness and gaps in defensive capability

Summary Takeaways

Enterprise Supply Chain Risk:
The Glass Worm attack marks a new height in supply chain sophistication. Security teams must respond immediately, with inventories, vetting, and restricted extension policies.
AI-Powered Toolchain Dangers:
Popular AI-based development tools are shipping with dozens of known vulnerabilities, with vendors slow or unwilling to patch—an urgent red flag for any security-conscious tech org.
AI as a Hacker's Superpower:
Survey data highlights the mounting fear and unpreparedness around AI-driven social engineering and deepfake attacks. The profession is facing a race to adapt defensive tactics as AI’s impact accelerates.

For further resources or to share tips, reach Jim Love via the Contact Us form at technewsday.com or .ca.

This summary captures the actionable insights and urgency in Jim Love’s coverage of AI-powered, supply chain, and social engineering threats—an essential listen for anyone grappling with the new realities of cybersecurity in 2026.

Loading summary

Transcript1 lines

[00:01]
A
Before we get started with today's show, I want to give a shout out to Brent Fowles, Director of Cybersecurity and Business Services at Western University in London, Ontario, Canada. I had the honor of presenting Brent with the CISO of the Year Award along with my co host Sean Guthrie, the President of the CIO association of Canada. CIO can these awards recognize leadership, seeing clearly through complexity, making bold choices and guiding teams through pressure and change? As one of the judges, I can say that these national awards are highly competitive and even being one of the finalist nominees is an honor. Congratulations Brent. And for those of you who might not know, CIOCAN is a peer run organization. It's run by CIOs and CISOs from across Canada. And for those whose Alphabet ends in a Z, it's a fantastic network for a ciso. Don't burn out and be on your own. Join up and support each other with peers like Brent OpenAI's Atlas browser faces early security concerns Glass Worm hides in Visual Studio Code extensions with over 35,000 downloads, AI IDEs ship with 94 known chromium bugs and exposing 1.8 million developers, and ISACA says AI driven attacks are now the top security fears for 2026. Given today's story, is it any wonder this is cybersecurity today? I'm your host Jim Love. Let's get into it. A new self spreading worm called Glass Worm has been discovered inside Visual Studio code extensions and one report said it's already been downloaded more than 35,000 times before being detected. Researchers at Coy Security found the malware hidden in seven extensions on the open VSX Marketplace and at least one in Microsoft's own VS Code Marketplace. They describe it as one of the most sophisticated supply chain attacks they've ever seen, and have posted a list of affected extensions on their website so organizations can check their Systems. What makes Glassworm so dangerous is how well it hides the malicious JavaScript is written using invisible Unicode characters, variation selectors that show up as blank space in the editor. To a developer it looks like empty lines, to scanners it looks like nothing, but to AI, it's executable code. Once installed, the malware steals developer credentials from GitHub, Git and NPM, drains funds from crypto wallets, turns developer machines into proxy servers, and installs hidden remote access tools for control. It even uses the Solana blockchain and a Google Calendar event for command and control, making it unusually difficult to shut down. Security teams and organizations where VS code is used should treat this as an active incident, you need an inventory of installed extensions and a comparison against that known affected list. You need to turn off automatic updates and review every extension across developer machines. Now, Koi is a vendor and undoubtedly wants to sell you some tools, but right now their site has the best list of affected extensions we've been able to find. And if anybody has a better verified source, please let me know. I'll post it on Monday's show. For the moment, the safest move is to block untrusted marketplaces and allow list only approved extensions until the threat is fully contained. Security researchers at AUX Security say the latest release of The AI powered IDEs cursor and windsurf are running on outdated components, including old versions of chromium and Google V8 JavaScript engine, leaving them vulnerable to more than 94 already patched security flaws. Together, these tools are used by an estimated 1.8 million developers worldwide. The researchers demonstrated an exploit using one of those bugs, an integer overflow flaw tracked as CVE2025 7656, showing that a malicious link could crash the editor and in real world conditions might allow for remote code execution. Now, despite the issue being responsibly disclosed on October 12, the risks remain bleeping. Computer reports that Cursor labeled the report out of scope, while Windsurf didn't respond at all. And that's a problem because the Electron based apps like these package a fixed Chromium build. Unless developers regularly update that embedded browser, they inherit every browser vulnerability that's been fixed since their last release. In this case, that means 94 of them, some of them critical. If your developers are using Cursor or Windsurf, put this on your incident list and press these vendors for patched builds. And if you hear that that's out of scope or you don't get an answer, maybe it's time to go looking for a new tool set. I'm a huge booster of moving forward towards an enterprise level of vibe coding, but we have to do it responsibly. AI is fast becoming a hacker's weapon of choice. A new global survey from Isaca finds that cybersecurity professionals are now seeing driven attacks as their top threat heading into 2026, overtaking ransomware and insider breaches for the first time. According to ISACA's 2026 Tech Trends and Priority Pulse poll, nearly 6 in 10 security audit and risk professionals say AI powered threats will keep them up at night next year 59% cited AI and deepfakes as their biggest concern, compared with 36% worried about delayed breach detection and 35% citing insider threats. And it's not just theoretical 63% said AI driven social engineering, from cloned voices and fake videos to automated phishing is now their biggest fear. Attackers are already using AI to craft convincing messages, mimic executives, and scale deception in ways human criminals never could. Only 13% of respondents said their organizations are very prepared to handle these risks. Nearly a third admitted they're not ready at all. The speed of AI adoption is outpacing the defenses meant to contain it, and the strain is starting to show, with 41% of professionals saying simply keeping up with AI's pace is their biggest personal challenge. But unfortunately, AI won't wait, and neither can defenders. The technology that's transforming our business is now transforming the threat landscape faster than most organizations can respond. And that's our show for today. You can reach me with tips, comments or constructive criticism@technewsday.com or CA. Just use the Contact Us form. I'm your host, Jim Love. Thanks for listening.