Jonathan Ng
Bio
Jonathan Ng is a Singapore-based AI safety researcher and engineer currently working as a Project Officer at the Singapore AI Safety Institute (AISI). He completed the SERI MATS 3.1 program (Spring 2023), where he worked with the Cadenza Labs team on extending the 'Discovering Latent Knowledge' paper by contributing to the EleutherAI/elk library, running large-scale hyperparameter sweeps, and conducting original experiments in LLM probing. He holds a BComp in Computer and Information Systems Security from the National University of Singapore (2017-2021). His research includes co-authoring the MACHIAVELLI benchmark (ICML 2023), which measures trade-offs between rewards and ethical behavior in AI agents, and the Catastrophic Cyber Capabilities Benchmark (3CB, AAAI 2025 Workshop). He is also a co-author of CCS-Lib, a Python package for eliciting latent knowledge from LLMs published in the Journal of Open Source Software. Beyond research, Jonathan has been a key organizer of AI safety capacity-building in Singapore, founding the Research Engineering Camp for Alignment Practitioners (RECAP), directing AI Safety Fellowships at NUS and NTU, and serving as an instructor at the ML4Good Singapore bootcamp. He received funding from the Long-Term Future Fund for his early ML alignment skill development and his SERI MATS project.
Links
- Personal Website
- -
- Twitter / X
- -
- LessWrong
- -
Grants
from Long-Term Future Fund
from Long-Term Future Fund
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Mar 22, 2026, 10:29 PM UTC
- Created
- Mar 20, 2026, 2:53 AM UTC