Skip to content

Exploit Development, Agentic Vulnerability Discovery, and Fake PoCs

Practical AI

Hosted by Daniel Whitenack

45m
Transcript

Original podcast

Where to listen

Summary

Written by a model from the transcript; not yet reviewed.

This episode explores the intersection of artificial intelligence and cybersecurity, focusing on exploit development, vulnerability discovery, and the rise of AI-generated proof-of-concept exploits. The guest, Spencer McIntyre, Director of Exploit Development at Volncheck, explains how AI models excel at identifying vulnerabilities in open-source web applications with logic flaws, but struggle with closed-source software requiring reverse engineering. The conversation covers how security teams use frameworks like Go Exploit alongside large language models to quicken their pace, while navigating the challenges of model guardrails, safety restrictions, and the surge of broken or fake proof-of-concept exploits appearing on GitHub.

  • Organizations should take major vulnerabilities seriously and use the window before full research is published to patch their systems. 13:09
  • Security teams can speed up exploit research by pairing structured frameworks with large language models to avoid building scaffolding from scratch. 18:07
  • AI models currently perform well at exploiting open-source web applications with logic flaws, but struggle with closed-source firmware and software. 19:24
  • Over half of the public proof-of-concept exploits found online are broken or illegitimate, largely due to an increase in AI-generated submissions. 30:51

Chapters

  1. 0:02
    Podcast Introduction and Host Overview

    Hosts Daniel and Chris introduce the episode and set the stage for the discussion.

  2. 1:09
    Defining Exploit, Vulnerability, and POC

    The hosts and Spencer clarify the distinctions between vulnerabilities, exploits, and proof‑of‑concepts.

  3. 2:04
    AI’s Impact on Vulnerability Discovery and Exploitation

    Discussion on how AI is accelerating vulnerability identification and the challenges it introduces.

  4. 2:35
    Exploit Development Background and Role at Volncheck

    Spencer explains his career path, current responsibilities, and the focus of Volncheck's Initial Access Team.

  5. 8:00
    Fake AI‑Generated POCs and the Rise of Illegitimate Proofs

    Exploration of the surge in AI‑generated POCs, their quality issues, and the impact on the security community.

  6. 21:22
    AI Guardrails and Legal Considerations for Researchers

    Spencer discusses the constraints of using commercial AI models for offensive security research and the legal implications.

  7. 34:27
    Future of Agentic Vulnerability Discovery and Exploitation

    Speculation on how agent swarms and AI will shape real‑time vulnerability hunting and exploitation.

  8. 41:08
    Closing Thoughts and Outlook for AI Security Research

    Spencer shares his optimistic view on the future of AI‑assisted security and the importance of speed.

Highlights

  • 7:07

    Defining Vulnerabilities Versus Exploitsinsight

    The guest explains that a vulnerability is the abstract flaw or potential for something bad to happen, whereas an exploit is a concrete implementation that leverages it to achieve attacker utility.

  • 9:39

    AI Excels at Open Sourceinsight

    Large language models do a good job identifying and exploiting logic flaws in open-source web applications, but struggle heavily with closed-source software that requires reverse engineering.

  • 22:40

    Stricter Guardrails Lead to Local Modelsstrategy

    As American proprietary models become more restrictive with safety guardrails, offensive security researchers are increasingly turning to local open-weight models like GLM.

  • 30:36

    Spike in Broken Proof of Conceptsdata

    Data from Volncheck shows that in 2026, the proportion of public proof-of-concept exploits that are broken or fail to meet quality standards tipped past fifty percent.

  • 35:16

    Agentic Swarms and Fuzz Farmsinsight

    The guest compares agentic research swarms to historic fuzz farms, noting that researchers are building custom harnesses and prompts rather than relying on off-the-shelf tools to find unique bugs.

  • 38:25

    Humans Still Needed for Impactdebate

    AI models are not yet as reliable as humans at driving exploits to achieve high-impact real-world utility like remote code execution.

Quotes

  • “A vulnerability is basically the potential for something bad to happen.”
    Unnamed speaker · 7:07
  • “A POC is kind of like somewhere in between.”
    Unnamed speaker · 27:57
  • “AI is not as reliable at exploiting them with the same impact as humans, because humans, first of all, they're going to be motivated for looking at the vulnerabilities.”
    Unnamed speaker · 38:25
  • Hostinger 15:35–16:41

About this episode

AI is changing cybersecurity, accelerating vulnerability research and exploit development while raising new questions about reliability, risk, and guardrails. Daniel and Chris talk with Spencer McIntyre, Director of Exploit Development at VulnCheck, about how agents are being used to discover vulnerabilities, why many proof-of-concept exploits (POCs) fail to deliver, and how initial access intelligence helps security teams prioritize real threats. They also explore the limitations of AI-assisted exploitation, the challenges of AI guardrails for security researchers, and what these developments mean for the future of threat intelligence. Featuring: Spencer McIntyre – LinkedIn https://www.linkedin.com/in/spencer-mcintyre-b48987194/Daniel Whitenack – Website https://www.datadan.io/, GitHub https://github.com/dwhitena, X https://x.com/dwhitenaChris Benson – Website https://chrisbenson.com/, LinkedIn https://www.linkedin.com/in/chrisbenson, Bluesky https://bsky.app/profile/chrisbenson.bsky.social, GitHub https://github.com/chrisbenson, X https://x.com/chrisbensonLinks: VulnCheck https://www.vulncheck.com/Go Exploit framework https://github.com/vulncheck-oss/go-exploitInitial Access Intelligence https://github.com/vulncheck-oss/initial-access-communitySponsors: Hostinger https://www.hostinger.com/in: Create your website today. Go to hostinger.com/practical and use practical for 10 percent off now. Resources and Events: Register for upcoming webinars here https://practicalai.fm/webinars! Prior Webinars from our partner Prediction Guard https://www.youtube.com/playlist?list=PLRpRdFmBfdF6boDgAD3nNaLm8Yx4KN2VtMidwest AI Summit 2026 https://midwestaisummit.com/Chris’s new book The Art of Swarming will be published soon, written to be the seminal foundational book for the fledgling swarming industry, defining the principles and architecture that enable autonomous agentic UxS swarms with emergent unified embodied intelligence.

Transcript

0:00
45:13

8,273 words · 3 speakers

Speaker names marked are inferred from speaking time and the show’s known hosts — they are not voice-verified.

Topics

  • Volncheck
  • Metasploit
  • OpenAI
  • GLM
  • Go Exploit
  • GitHub
  • Prediction Guard
  • Ruby
  • Practical AI Podcast
  • Rails

Page updated · transcript and summary are machine-generated