I'm Ma'alona Mafaufau — AI safety researcher, technical consultant, and full stack developer based in Auckland, New Zealand.

I came to this work through an unconventional path — statistics, actuarial modelling, data analytics, cybersecurity, and now software development. That winding journey means I look at problems from multiple angles: not just the technical, but the creative, cultural, and strategic dimensions too.

What I Do:‍ ‍

I build purpose-built digital platforms for organisations doing meaningful work in Aotearoa. From Māori outcomes tracking in infrastructure to AI-powered creative operations — I handle the full lifecycle from research and architecture through to production build. My background in data, security, and AI safety means the platforms I build are designed with data sovereignty, privacy, and long-term resilience from day one.

What I’ve Discovered:‍ ‍

I placed 3rd in the Palisade Research AI Misalignment Bounty, demonstrating reproducible misalignment behaviours in frontier AI models including GPT-5 and o3. That work was published on arxiv and released on huggingface.

I built the G1 Alignment Experiment — an embodied AI alignment platform using MuJoCo physics simulation that tests LLMs controlling a humanoid robot under resource pressure. The research revealed that models violate safety constraints at specific pressure thresholds, with four distinct failure modes: rationalisation, no ethical anchor, explicit override, and accountability avoidance.