About

Future Ethics is an AI research lab in San Francisco. We focus on procurement and multilingual safety and bias audits and evaluations.

Who We Are

Adrianna Tan runs Future Ethics, an AI lab focused on creating procurement and multilingual safety standards in San Francisco. She was AI Red Team Lead at Humane Intelligence (founded by Dr. Rumman Chowdhury), leading large-scale evaluations like a crowdsourced medical chatbot pilot with 200+ healthcare professionals that found 800+ vulnerabilities. She was also the project lead on IMDA Singapore’s world-first AI Safety project in 9 countries and 9 languages.

Before AI safety, she built products at San Francisco Digital Services and PlanetScale, shipping Covid-19 vaccine search systems and cloud native database services to millions of users.

Niall Saville is a strategic advisor who has worked on the socioeconomic impacts of new technologies throughout Asia, Africa, Europe and the Middle East. Most recently he led economic prosperity workstreams for the Tony Blair Institute in Indonesia. He also supported the Indonesian Ministry of Communications and Digital Affairs with their strategic approach to new and frontier technologies, including AI. Prior to TBI Niall worked as a strategic advisor for the UN, as an Associate Partner for Dalberg, and for Monitor Group. He has lived and worked in Kenya, Tanzania, Rwanda, and South Africa as well as Vietnam, Indonesia, Singapore, the UAE and Saudi Arabia.

What We Do

• Fully managed pre-procurement AI safety and bias testing
• Multilingual AI safety and bias audits
• Large-scale crowdsourced testing

AI Red Team

• We have expertise running real world AI red teams in multiple domains
• In 2025, we led the Department of Defense's military medicine LLMs red team exercise
• In 2024, we led the Singapore government's 9-country, 8-languages red team exercise

Multilingual Evals

• We have led AI red teams in English, and 10 other languages
• We have an active working group with NLP and ML researchers to work on how to improve tokenization in other languages

Pioneers of hybrid testing

• We are equally comfortable building a multi-agent LLMs-as-judge workflow, as we are with running a fully manual testing suite with human experts

Community of practice

We want to collaborate with people and organizations who want to explore new business models centered on people, labor and ethics.