Find your fit in AI safety research

MATS fellows work across seven tracks spanning technical research, governance, biosecurity, strategy, theory, security, and field-building. Explore each track to learn what fellows work on, who the track is for, and how to apply.

Winter 2027 Tracks

View MATS Winter 2027 tracks and their streams.

创业与领域建设

AI 安全需要快速扩大规模,而瓶颈正日益转向组织建设。本方向面向希望创办 AI 安全机构或项目的创业者、领域建设者和积极主动的通才,并由相关领域的创业者、现任 CEO 与项目负责人提供指导。

View track
Juniper Ventures
Dewi Erwan (BlueDot Impact)
Rosie Campbell (Eleos)
Ross Matican & Mike McCormick (Halcyon Futures)
Alex Holness-Tofts (Resolution)

实证研究

通过机器学习实验开展实践研究,以理解并提升模型安全性,涵盖 AI 控制、可解释性、可扩展监督、评估、红队测试和鲁棒性。

View track
David Africa
Michael Chen
David Lindner
Patrick Butlin
Oliver Crook
Maksym Andriushchenko
Transluce
SecureBio AI
Team Shard
Redwood Research
Apollo Research - Monitors
Tomek Korbak
Sunishchal Dev
Fourth Eon Biosecurity
LILA (David Peinador Veiga)
Anthropic
Alfie Lamerton & Joe Kwon
Truthful AI
Roger Grosse
Caspar Oesterheld (Redwood, conceptual reasoning capabilities)
Neel Nanda
Joshua Engels
Victoria Krakovna
Apollo Research - Science of Scheming
Daniel Kang

政策与治理

研究先进 AI 应该如何治理,涵盖治理机制、监管与制度分析,以及让治理措施得以执行的技术系统。

View track
Aidan O'Gara
Dave Banerjee
Safe AI Forum
Matthew Gentzel
Mauricio Baker, Anjay Friedman
McKenna Fitzgerald
Michael Chen
Patricia Paskov & Miles Brundage (AVERI)
TAIGR
Luke Drago, Rudolf Laine
FAI AI Team
Sambhav Maheshwari, Jan Wehner
Peter Wildeford

Biosecurity

Research on catastrophic biological risk in a world being reshaped by advanced AI. Spans pathogen detection, medical countermeasures, synthesis screening, physical biodefense, threat modeling, and red-teaming biological AI for dangerous capabilities.

View track
Oliver Crook
Active Site
SecureBio AI
Damon Binder
Sebastian Oehm, Askar Kleefeldt
Aman Patel, Adin Richards
Sunishchal Dev
Fourth Eon Biosecurity
LILA (David Peinador Veiga)
Sentinel Bio - Toby Webster - AIxBio

战略与预测

研究 AI 发展可能如何演变,以及这对长期安全意味着什么。涵盖发展时间线、能力跃迁动态、风险建模和 AI 发展轨迹的战略分析。

View track
Dave Banerjee
Civilizational Alignment
AI Futures Project
Damon Binder
Luke Drago, Rudolf Laine

Theory

Foundational research on the mathematical and philosophical principles underlying agency, alignment, and safe reasoning in advanced AI systems.

View track
Abram Demski
Caspar Oesterheld (Redwood, conceptual reasoning capabilities)
Alignment Research Center (ARC)
Patrick Butlin
Richard Ngo

系统安全

研究先进 AI 所依赖基础设施的软件与硬件安全,包括侧信道分析、集群安全、模型权重保护和物理层验证。

View track
Gabriel Kulp
Lisa Thiergart, Luis Cosio, Guy (SL5 Task Force)
Mauricio Baker, Anjay Friedman
Daniel Kang