Researcher

Joe Benton

Embedded assessment of AI risks at METR

Benton studies how to assess and reduce risks from advanced AI. At Anthropic he led work on scalable oversight—ways to supervise systems that may outperform people—and researched tests for deceptive or misaligned behavior; he joined METR to develop independent, embedded risk assessments.

See something inaccurate or outdated?

Background

Current role

Embedded assessment of AI risks Aug 2026-present

Previously

Nov 2023-Aug 2026Member of Technical Staff; Manager, Scalable Oversight,

Fall 2023-Fall 2023Technical Staff, UK Frontier AI Taskforce,

AI Security Institute

Winter 2022-2023Machine Learning Researcher,

Redwood Research

Education

University of Oxford

University of Cambridge

About this data

Work and education history is primarily focused on AI-relevant roles and may not be comprehensive.

Last editorial review: September 6, 2026

News 1

moveJoe Benton leaves Anthropic for METR

The former Scalable Oversight manager has joined METR to work on independent, embedded assessments of AI risks.

September 2, 2026

Citation Trend

1 snapshots

2,853citations as of Sep 6, '26

For today’s citation count and citations per paper, visit Google Scholar.