2026
Papers · Patents · Writing
Research.
Understanding the trust boundaries of AI systems, from prompt injection to defensive applications.
Google Scholar profilePapers & publications
2025
Prompt Injection 2.0: Hybrid AI Threats
Read on arXiv2024
AI Ethics by Design: Implementing Customizable Guardrails for Responsible AI Development
Read on arXiv2023
Defensive AI: Experimental Study
Find on Google Scholar2022
Evaluating the Susceptibility of Pre-Trained Language Models via Handcrafted Adversarial Examples
Read on arXiv2021
An Observational Assessment of CTI Standards for Blue Teams
Read proceedings (PDF)Issued patents
Prompt injection mitigation
Mitigation for prompt injection in A.I. models capable of accepting text input
Inventors: Jonathan Cefalu, Jeremy Charles McHugh, and Ron Heichman.
Selected writing
ToolJack: Hijacking an AI Agent’s Perception Through Bridge Protocol Exploitation
Research into agent perception under a compromised local bridge, with the local-access prerequisite and observed limits described.
Exploiting Along the Promptware Kill Chain: AI Red Teaming
An examination of prompt injection and agent workflows through a security-testing lens.
5 Years of Defending AI and the Vulnerability Class That Reshaped the Industry
A retrospective on the evolution of AI security and prompt injection.