Jack Payne

AI Safety Researcher

ML Engineer and AI Safety Research Fellow based in Sydney. I think AI is important but we're a long way off having it be safe.

Currently researching AI time-horizons for offensive cybersecurity tasks as an independent researcher, and building interpretability agents as a SPAR fellow.

Work

Independent Research Engineer
AI time-horizons for offensive cybersecurity (with Sean Peters)
2025
SPAR Research Fellow
Building interpretability agents
2025
Ampelic AI
ML Engineer - building a POC for seed round
2024
TARA & Oxford ARBOx
AI Safety programs
2024

Projects

AI Risk Monitor
Safety index for frontier models
Ongoing
SpotTheSlop
AI vs human media game
2024
Crosscoder Model-Diffing
Stage-wise interpretability research
2024
Deepseek CoT Model-Diffing
Base vs distilled CoT models
2024
Subliminal Learning
Investigating properties of subliminal learning
2024

Goals

Help ensure AI benefits humanity/7 Marathon Majors (0/7)/Release an EP/Learn piano

Personal

I like songs that feel like a transitional time, DJing & producing music, and running. I dislike giving AI autonomy haphazardly.

Reading