♛Arshi Chadha

AI security researcher · Senior PSIRT Engineer, Zscaler

I've spent 8+ years in security. These days I run product security incident response at Zscaler, and the rest of my time goes into studying how AI systems get attacked, and what to do after they do. Most AI security work stops at the jailbreak. Mine starts there.

Breaking Models

In 2026 I single-handedly started Breaking Models, an AI security meetup in the Bay Area. The first session was ten people around a table at Hacker Dojo. Now it's a full house every month: engineers, researchers, founders, and security leaders who show up to break AI systems together. Prompt injection, RAG poisoning, embedding attacks, live demos, real CVEs.

Talks first, then open floor. Beginner friendly, discussion heavy, and no vendor pitches, ever. Want to attend or give a talk? Write to aisecurityinbay@proton.me.

Research, reviews, rabbit holes

If you're working on AI security and want a collaborator, a co-author, or a second pair of eyes on a draft, my inbox is open. I review papers, CFP submissions, and half-finished exploit ideas with equal enthusiasm. The best emails I get start with "this might be a dumb idea, but".

A few things about me

• Co-Lead of LLM08 (Vector and Embedding Weaknesses) in the OWASP Top 10 for LLM Applications, 2026
• Speaker at BSides Las Vegas 2026: "After the Jailbreak", a PSIRT playbook for agentic AI
• MS in Information Security from Carnegie Mellon University
• Co-inventor on a granted US patent at Zscaler

Off the clock

Chess, mostly. Also old Bollywood soundtracks, and chai strong enough to wake a night-shift SOC analyst. Chess challenges welcome, in games and in research.

Say hi

For research, Breaking Models, speaking invites, or a game of chess: arshilife16@gmail.com · LinkedIn · GitHub