Yash Maurya

Research Engineer

I'm a Research Engineer at Scale AI where I lead LLM red teaming and adversarial testing for frontier AI labs. My work centers on finding privacy and safety vulnerabilities in AI systems before they reach production.

Currently working on:

  • Red teaming safety and privacy capabilities for frontier AI laboratories
  • Building automated frameworks to evaluate enterprise LLMs against content policies
  • Developing attack taxonomies and defenses for large-scale ML systems

During my time at CMU's Privacy Engineering Program, I conducted research under the mentorship of leading privacy experts including Professors - Hana Habib, Lorrie Cranor, Norman Sadeh, and Virginia Smith. My work focused on developing practical privacy solutions through research and industry collaborations with companies like Meta and PwC, resulting in publications at PEPR, SOUPS, ICLR, SaTML and earning recognition as an IAPP Westin Scholar.

During Summer 2024, I worked as an AI Governance intern at BNY AI Hub, helping develop Eliza—BNY's AI platform used by 15,000+ people. I implemented LLM safety measures, built benchmarking pipelines, and contributed to governance frameworks. This project was featured in Fortune magazine.

Research Interests

Privacy-Preserving ML, Differential Privacy, Federated Learning, and Responsible AI

Here's my resume in case you need it.

Want to chat? Send me an email or text on LinkedIn!

Yash Maurya
Privacy is not something that I'm merely entitled to, it's an absolute prerequisite.

Recent News

Research

* indicates equal contribution

2025

MoReBench: Evaluating Procedural and Pluralistic Moral Reasoning in Language Models, More than Outcomes
[website]
Yu Ying Chiu*, Michael S. Lee*, Rachel Calcott, Brandon Handoko, Paul de Font-Reaulx, Paula Rodriguez, Chen Bo Calvin Zhang, Ziwen Han†, Udari Madhushani Sehwag, Yash Maurya, Christina Q. Knight, Harry R. Lloyd, Florence Bacus, Mantas Mazeika, Bing Liu, Yejin Choi, Mitchell L. Gordon, Sydney Levine
arXiv preprint arXiv:2510.16380; under submission to ICLR 2026
TL;DR: Introduces MoReBench, a benchmark of 1,000 moral scenarios and 23k+ rubric criteria to evaluate how language models reason about moral dilemmas, emphasizing pluralistic and process-focused moral reasoning over just final outcomes.
Making Privacy-Preserving AI Accessible: A Practitioner-Oriented Framework
[program]
Yash Maurya, Ibrahim Mohamed Anis Chhaya, Hana Habib
IEEE Symposium on Privacy Expectations (ISoPE) 2025 & SUPA 2025 Workshop on Societal & User-Centered Privacy in AI
TL;DR: Practitioner-oriented framework that organizes concrete ML privacy mitigations, tools, and design patterns across the ML lifecycle to help teams operationalize privacy-preserving AI in real-world deployments.
When Privacy Guarantees Meet Pre-trained LLMs: A Case Study in Synthetic Data
Yash Maurya*, Aman Priyanshu*
2025 USENIX Conference on Privacy Engineering Practice and Respect (PEPR'25)
TL;DR: Case study evaluating how privacy guarantees behave when using pre-trained LLMs to generate synthetic data, highlighting practical leakage risks.

2024

Beyond the Accept Button: How Information and Control Shape Data Sharing and AI Engagement
Ibrahim Chhaya*, Yash Maurya*, Zuofei Hong*, Limin Ge*
Sponsored by Meta - MSIT-PE Capstone Report 2024 (Advisor: Professor Lorrie Cranor)
TL;DR: Studied how consent flow design impacts AI engagement and data sharing, examining length and control options in social media data sharing.
Position: LLM Unlearning Benchmarks are Weak Measures of Progress
Pratiksha Thaker, Shengyuan Hu, Neil Kale, Yash Maurya, Zhiwei Steven Wu, Virginia Smith
IEEE Conference on Secure and Trustworthy Machine Learning (SaTML), 2024
TL;DR: Discussion about the current state and limitations of LLM unlearning benchmarks with a focus on forget/retain set methods
Designing a Benefit Assessment Protocol for AI Systems
Rachel Kim*, Yash Maurya*, Goutam Mukku*
Course Project for Responsible AI Course(10-735) at CMU (Advisor: Professor Hoda Heidari)
TL;DR: A structured protocol for systematically assessing AI benefits to enable more comprehensive AI evaluation.
Unified Locational Differential Privacy Framework
Aman Priyanshu*, Yash Maurya*, Suriya Ganesh*, Vy Tran*
arXiv preprint arXiv:2405.03903 (Advisor: Professor Hana Habib)
TL;DR: A privacy framework for aggregating sensitive location-based data while protecting individual privacy through differential privacy mechanisms.
AI Governance and Accountability: An Analysis of Anthropic's Claude
Aman Priyanshu*, Yash Maurya*, Zuofei Hong*
arXiv preprint arXiv:2407.01557 (Advisor: Professor Norman Sadeh)
TL;DR: Case study on Anthropic's Claude using AI governance and accountability frameworks, examining compliance with NIST and EU AI Act standards.
Guardrail baselines for unlearning in LLMs
Pratiksha Thaker, Yash Maurya, Shengyuan Hu, Zhiwei Steven Wu, Virginia Smith
ICLR 2024 Workshop on Secure and Trustworthy Large Language Models
TL;DR: Simple guardrails (prompting and filtering) match finetuning's effectiveness for unlearning in LLMs, challenging current evaluation metrics.
Design and Evaluation of the UsersFirst Privacy Notice and Choice Threat Analysis Taxonomy
[poster]
Xinran Alexandra Li*, Yu-Ju Yang*, Yash Maurya*, Tian Wang*, Hana Habib, Norman Sadeh, Lorrie Faith Cranor
Twentieth Symposium on Usable Privacy and Security (SOUPS 2024 Posters) & SOUPS 2024 Societal & User-Centered Privacy in AI Workshop (SUPA 2024)
TL;DR: UsersFirst taxonomy outperforms LINDDUN PRO in detecting privacy notice and choice threats in user study.
UsersFirst: A User-Centric Privacy Threat Modeling Framework for Notice and Choice
[poster]
Tian Wang*, Xinran Alexandra Li*, Miguel Rivera-Lanas*, Yash Maurya*, Hana Habib, Lorrie Faith Cranor, Norman Sadeh
Twentieth Symposium on Usable Privacy and Security (SOUPS 2024 Posters) & SOUPS 2024 Workshop on Privacy Threat Modeling (WPTM 2024)
TL;DR: UsersFirst: A user-centric framework for identifying and mitigating privacy notice and choice threats, extending beyond LINDDUN
Through the Lens of LLMs: Unveiling Differential Privacy Challenges
Aman Priyanshu*, Yash Maurya*, Vy Tran*
2024 USENIX Conference on Privacy Engineering Practice and Respect(PEPR'24)
TL;DR: LLMs demonstrate stronger privacy attacks on Google's Topics API, bypassing differential privacy safeguards.
Is it Worth Storing Historical Gradients?
Joong Ho Choi*, Yingxin Liu*, Yash Maurya*
Course Project for Federated and Collaborative Learning Course(10-719) at CMU (Advisor: Professor Virginia Smith)
TL;DR: Current weights beat historical gradients for detecting FL attacks, saving storage and enhancing privacy.

2022

Federated Learning for Colorectal Cancer Prediction
Yash Maurya*, Prahaladh Chandrahasan*, G Poornalatha
2022 IEEE 3rd Global Conference for Advancement in Technology (GCAT), 1-5
TL;DR: Federated learning enables privacy-preserving colorectal cancer prediction across hospitals with centralized-level accuracy

2021

Improved variants of Score-CAM via Smoothing and Integrating
[poster only]
Rakshit Naidu, Haofan Wang, Soumya Snigdha Kundu, Ankita Ghosh, Yash Maurya, Shamanth R Nayak K, Joy Michael
Responsible Computer Vision (RCV) Workshop at CVPR 2021
TL;DR: Slightly modified version of IS-CAM (described below)

2020

IS-CAM: Integrated Score-CAM for axiomatic-based explanations
Rakshit Naidu, Ankita Ghosh, Yash Maurya, Shamanth R Nayak K, Soumya Snigdha Kundu
arXiv preprint arXiv:2010.03023
TL;DR: Enhanced CNN interpretability through IS-CAM, integrating Score-CAM to produce sharper attribution maps