The Terminology Problem Causing Security Teams Real Risks
Jailbreaks target the model's safety training; prompt injection hijacks application trust boundaries. Conflating them leads to defenses that miss your actual threat surface.
About Me
Ariel Fogel
I’m a Software Engineer and Researcher with extensive experience across research and development roles. I’m passionate about improving how people learn and advancing cybersecurity.
Cold, Medium, and Hot Takes
View all posts »
What Makes Indirect Prompt Injections Attacks Succeed
Anatomy of an Indirect Prompt Injection
Sane Python Debugging in VSCode
A path towards a more reasonable dev experience
Mastering Stream Detection: Using MutationObserver to Track LLM Responses in Real-Time
A vanilla JavaScript solution for detecting when streaming responses from Large Language Models have completed
About me
Professionally, I spend time thinking about web development, AI security, learning, and design. This space houses my thoughts on these topics, among others.
When I'm not working, I'm practicing yoga, cooking, or playing jazz.
Research & Talks
Explore my academic publications and conference presentations
Publications
Peer-reviewed research on learning sciences, AI security, and educational technology.
Presentations
Conference talks, workshops, and congressional briefings on AI literacy, security, and web development.
Experience
Founding Engineer and Researcher
Pillar Security
Tel-Aviv, Dec 2023 - Present
Core founding engineer building the product from 0 to 1, contributing to a $9 million seed round.
- • Architected full-stack AISPM (AI Security Posture Management) and EDR platform
- • Built AI-powered threat modeling and red teaming suite for LLMs
- • Trained and deployed production classifier models for attack detection
- • Published security research on AI supply chain vulnerabilities
Software Engineer
The Agency of Learning
Remote, 2023
I design and implement open source collaboration tools for developers using Ruby on Rails 7, Hotwire, and Tailwind. Responsibilities include design (wireframing), end-to-end implementations of features, writing design documents and other documentation, setting up CI/CD, and code review.
Technical Founder
Edurrhaphy LLC
Remote, 2021 - 2023
Using Ruby on Rails 7, I built a fully functional flashcard application to teach surgeons anatomy. Platform includes both authoring tools to create content in addition to a flashcard game.
Researcher
The MAGIC lab @ UW-Madison
Madison, WI, 2020 - 2022
I built a serverless augmented reality game used as the basis for four publications. Architected and managed a team of research interns contributing to the development of the platform, including performing code review on contributions
Data Engineer
MasterClass
2018-2019
I designed analytical views reflecting business processes to be used for the company’s OLAP cube infrastructure. Additional responsibilities included building Looker dashboards to communicate critical business metrics to stakeholders througout the organization
Software Engineer
Computational Approaches to Human Learning Lab @ UC-Berkeley
2017-2018
I expanded functionality of a data visualization tool enabling researchers to explore high dimensional datasets using d3.js. Managed a team of programmers building out a course recommendation system used by hundreds of students at UC-Berkeley.
Software Engineer
Bloc
2015-2017
Working with the CFO, I architected and led the planning, implementation, TDDing, and deployment of a critical upgrade to the company’s payments/billing infrastructure. I also worked with designers, product managers, and customer success teams to implement various student-facing and customer support features
Presentations
LLM Backdoors in Plain Sight: The Threat of Poisoned Templates
BSidesTLV 2025
Tel-Aviv, Israel, December 2025
Presented the same research to the Israeli security community at one of the largest security conferences in the region.
LLM Backdoors in Plain Sight: The Threat of Poisoned Templates
OWASP Global AppSec 2025
Washington, DC, November 2025
Presented research on supply chain vulnerabilities in AI systems, demonstrating how malicious actors can poison open-source LLM templates to execute backdoor attacks. Watch the recording →
AI Literacy 101
Congressional Briefings
Washington, DC, November 2025
Briefed members of the Joint Economic Committee and House Committee on Financial Services on AI fundamentals for policymakers. View slides →
Education
Master of Arts in Learning Sciences
UW Madison, Madison, WI
2019 - 2022
Dogtags
Dev Bootcamp, San Francisco, CA
2014, "I must create my own system or be enslaved by another."
Bachelor of Arts in Behavioral Economics; Music
Muhlenberg College, Allentown, PA
2007 - 2011