Intro to AI Safety Talk
About the talk
We are building the AI safety community in the Philippines, and this starts with the fundamentals.
The Philippines imports AI systems built by companies in other countries. We don't control how these models are trained, but we deal with the consequences when they're deployed here. Biased hiring algorithms, election deepfakes, content moderation that doesn't understand Philippine context — all of these are happening now.
Most AI safety work focuses on the companies building frontier models. Less attention goes to how countries should govern the AI systems they import and deploy. For the Philippines, the basic questions haven't been systematically answered: which risks matter most here, what governance actually looks like in practice, and how people can contribute to solving these problems.
This talk explains how AI systems work, what the governance challenges are globally and locally, and where Filipinos can contribute to solving them.
About the speakers
Kyle Reynoso was a Research Fellow at Pivotal Research, working on fundamental interpretability approaches to better understand and ensure the safe behavior of large language models. Previously, he was the Learning Director at WhiteBox Research, where he designed a mechanistic interpretability curriculum and mentored fellows in the Philippines and across APAC. He graduated with a BS in Computer Science from UP Diliman in 2023.
Lexley Villasis is the President of AI Safety Diliman. She is also a researcher at AI Safety Camp under Ananthi Al Ramiah, Ph.D., where they are looking into how individuals and communities can exercise agency over AI deployment. Previously, she was the Outreach Director of Condor Global SEA, an AI Safety camp for talented Southeast Asians held in Singapore. She is interested in Chinese AI policy and US-China dynamics on AI.
Danielle Castillo is an independent researcher working on the dynamics of the Global AI divide and Global South AI governance. She has participated in lobbying efforts for AI governance in PH Congress and is currently working on a publication on the same area of impact with Dr. Caroline Schoepf from UPD Sociology.
About AI Safety Diliman
AI Safety Diliman is a new organization at UP Diliman focused on training early-career professionals who can contribute to AI alignment research or governance work. We're supported by Kairos, an AI safety field-building nonprofit, and fiscally sponsored by Berkeley Existential Risk Initiative.
Express interest in our fellowship program at this link and learn more about AI Safety Diliman at https://aisafetydiliman.com.