The Rat Dev ReMix: Persona Implantation, Drift, and LLM Identity Attacks
Details
Just went you thought we were out, we are so back in!!! Just when we thought logistics were solved, they get so much more complex, don’t they? But don’t fret dear friends, our also dear friends in security at Reclaimer's Hall have offered us safe quarter to hold our first meeting of 2026, where https://lnkd.in/eCjMR_mi will be presenting us on his work, "The Rat Dev ReMix: Persona Implantation, Drift, and LLM Identity Attacks Through an OWASP Lens"
So please join us March 31, 7-9pm, at Reclaimers Hall (Ex Machina Parlor)
7-710 - Find your seats
710 - 715 - hello from Reclaimers Hall
715 - 730 - Chapter Updates (aka wtf is going on?)
730-815 - Speaker
815-830 - Questions/discussion
830-9 hang out, wrap up
Lite refreshments in the form of pizza and stuff will be served. Location info at the bottom. Note: This is an open meeting, but space is somewhat limited. I have no way of efficiently doing RSVP right now so just come out and we’ll make it work. There’s always a few pubs around the corner in our fair city.
Also, if you haven't already, follow us on LinkedIn for future updates:
https://www.linkedin.com/company/owasp-philly/
Enough from me, let’s hear it from Steve!
Brief Bio
Stephen Pote is a security practitioner focused on AI and LLM security, with experience in malware research, security tooling, and adversarial testing. His work examines how alignment, content policy, and safety controls in large language models fail under real-world adversarial pressure. Originally developed for PumpCon, this OWASP ReMix reframes that research for the AppSec and defender community, with the “Rat Dev” persona occasionally appearing as a co-speaker to illustrate attacker mindset and failure modes.
Brief Abstract
This talk is an OWASP-focused ReMix of a PumpCon presentation exploring how modern LLMs break under adversarial conditions. Topics include prompt injection, persona drift, gradient-style jailbreak techniques, and how these map directly to the OWASP Top 10 for LLM Applications. The session emphasizes practical threat modeling, red-team insights, and defensive considerations for organizations deploying AI-enabled systems.
Contact / Socials
[stephen.pote@internationalsos.com](mailto:stephen.pote@internationalsos.com)
[mause.konig@outlook.com](mailto:mause.konig@outlook.com)
https://lnkd.in/eCjMR_mi
7-9pm
Reclaimers Hall (Ex Machina Parlor)
319 N 11th Street, Philadelphia PA 19107
Room 5L (5th floor)