This week we follow-up on a story from last week's episode about the rise of Agentic Threat Actors (ATAs). Christina & Rory chat with Michael Clark, Senior Director of Threat Research at Sysdig about their discovery of JADEPUFFER, an ATA considered to be the first instance of agentic ransomware - an attack where a Large Language Model (LLM) is on the other end of the connection rather than a human operator. By exploiting the Langflow platform, the agent then autonomously attempted to perform automated database extortion against MySQL instances. But did it achieve its objectives?

Leave us a message!

Want to leave some feedback and suggestions but would prefer not to write an email? You can leave us a voicemail (90 seconds max). We will share any that are not too spicy in the following episode.

SpeakPipe link - https://www.speakpipe.com/Cyber_Sidekicks

Or, send us an email:

• Christina Richmond – christina@richmondadvisorygroup.com
• Rory Duncan – rory@richmondadvisorygroup.com

Subscribe to our Newsletter - "Signal, not noise" Our monthly newsletter is free to subscribers! Sign-up now at Richmond Advisory Group.

Technology we use

• Podcast Recording Platform – Cleanfeed.com
• Cyber Sidekicks Show Host – Podbean.com
• Edited & mastered in GarageBand

In this week's episode, Christina & Rory talk with guest Heath Renfrow, co-founder of breach recovery specialist Fenix24. Warning: It's a sobering look at the "cleanup" side of cybersecurity, emphasizing that most organizations are fundamentally unprepared for the complexity of full system restoration after a major attack. From the failure of backups and the hidden cost of business interruption, the discussion concludes the need for resiliency to be the top security control in every organization.

NEWS

JADEPUFFER: Agentic ransomware for automated database extortion

Startup sues Palo Alto Networks' Koi Security, saying an AI-hallucinated report falsely linked it to Chinese espionage

Leave us a message!

Want to leave some feedback and suggestions but would prefer not to write an email? You can leave us a voicemail (90 seconds max). We will share any that are not too spicy in the following episode.

SpeakPipe link - https://www.speakpipe.com/Cyber_Sidekicks

Or, send us an email:

• Christina Richmond – christina@richmondadvisorygroup.com
• Rory Duncan – rory@richmondadvisorygroup.com

Subscribe to our Newsletter - "Signal, not noise" Our monthly newsletter is free to subscribers! Sign-up now at Richmond Advisory Group.

Technology we use

• Podcast Recording Platform – Cleanfeed.com
• Cyber Sidekicks Show Host – Podbean.com
• Edited & mastered in GarageBand

In this week's episode, guest Jayesh Kamat, Global Competency Leader in Application Security at IBM, chats with Christina about the profound impact of AI frontier models on cybersecurity and the necessity of shifting toward an "AI-native" security posture.

With machines now generating code at an exponential rate through agents like "IBM Bob," traditional application security (AppSec) methods are no longer sufficient. Kamat advocates for AI-native security, which integrates protection directly into the development lifecycle:

• Secure by Design: Security requirements, such as whitelisted libraries and functions, should be defined in the specifications before any code is generated.
• Agent Verification: AI agents should be tasked with reviewing their own code, verifying its security, and explaining their "thought process" for human oversight.
• The Security Harness: This new approach involves using a specialized "harness" to deploy frontier models to identify vulnerabilities and map exploit paths.
• Once a path is identified, other agents can auto-remediate the code or update firewall controls in real-time

Ultimately, Kamat views this as a "moment under the sun" for application security, which has shifted from a neglected back-office function to a primary focus of enterprise strategy

Leave us a message!

Want to leave some feedback and suggestions but would prefer not to write an email? You can leave us a voicemail (90 seconds max). We will share any that are not too spicy in the following episode.

SpeakPipe link - https://www.speakpipe.com/Cyber_Sidekicks

Or, send us an email:

• Christina Richmond – christina@richmondadvisorygroup.com
• Rory Duncan – rory@richmondadvisorygroup.com

Subscribe to our Newsletter - "Signal, not noise" Our monthly newsletter is free to subscribers! Sign-up now at Richmond Advisory Group.

Technology we use

• Podcast Recording Platform – Cleanfeed.com
• Cyber Sidekicks Show Host – Podbean.com
• Edited & mastered in GarageBand