17 papers
Manifold of Failure: Behavioral Attraction Basins in Language Models
Sarthak Munshi, Manish Bhatt, Vineeth Sai Narajala +4
While prior work has focused on projecting adversarial examples back onto the manifold of natural data to restore safety, we argue that a comprehensive understanding of AI safety r…
The Defense Trilemma: Why Prompt Injection Defense Wrappers Fail?
Manish Bhatt, Sarthak Munshi, Vineeth Sai Narajala +6
We prove that no continuous, utility-preserving wrapper defense-a function that preprocesses inputs before the model sees them-can make all outputs strictly safe for a…
Adversarial Hubness Detector: Detecting Hubness Poisoning in Retrieval-Augmented Generation Systems
Idan Habler, Vineeth Sai Narajala, Stav Koren +2
Retrieval-Augmented Generation (RAG) systems are essential to contemporary AI applications, allowing large language models to obtain external knowledge via vector similarity search…
From Tool Orchestration to Code Execution: A Study of MCP Design Choices
Yuval Felendler, Parth A. Gandhi, Idan Habler +2
Model Context Protocols (MCPs) provide a unified platform for agent systems to discover, select, and orchestrate tools across heterogeneous execution environments. As MCP-based sys…
Large Empirical Case Study: Go-Explore adapted for AI Red Team Testing
Manish Bhatt, Adrian Wood, Idan Habler +1
Production LLM agents with tool-using capabilities require security testing despite their safety training. We adapt Go-Explore to evaluate GPT-4o-mini across 28 experimental runs s…
MAIF: Enforcing AI Trust and Provenance with an Artifact-Centric Agentic Paradigm
Vineeth Sai Narajala, Manish Bhatt, Idan Habler +2
The AI trustworthiness crisis threatens to derail the artificial intelligence revolution, with regulatory barriers, security vulnerabilities, and accountability gaps preventing dep…