AIA
An auditor for local AI agent deployments. Maps trust chains and credential scopes to help identify shared identities, broad permissions, and impersonation risks.
View sourceAI security · Research · Pittsburgh
Jeremy
McHugh, D.Sc.
AI security researcher & builder.
I study how AI systems fail and build tools to make agents, identities, and everyday software more trustworthy.
Selected work
From research to useful tools.
Super AI Markets
A public shopping environment for evaluating AI shopping agents and agentic browsers, including how they respond to untrusted content.
Visit environmentOmnia OS
An email client that groups messages by company, isolates unverified senders, and offers focused catch-up and unsubscribe workflows.
Visit websiteResearch & writing
A closer look at AI security.
The Anatomy of a Prompt Injection: A Component Model for Structured Analysis
A structured model for examining the components of prompt injection.
ToolJack: Hijacking an AI Agent’s Perception Through Bridge Protocol Exploitation
Research into how a compromised local bridge can change what an agent perceives.
Connect
Let’s compare notes.
For research discussions, useful tools, and collaboration.
Get in touch