Independent project. Not a U.S. government website.

Support our project (opens Buy Me a Coffee)
USASI

Samuel Marks

Leads the Cognitive Oversight subteam at Anthropic; first author of a study on auditing hidden objectives

  • Member of Technical Staff; leads the Cognitive Oversight subteam of the Alignment Science team, Anthropic124

Samuel Marks is a Member of Technical Staff at Anthropic and leads the Cognitive Oversight subteam of its Alignment Science team, which aims to oversee AI systems by examining the cognitive processes behind their behavior rather than only their inputs and outputs. Marks is first author and a core research contributor of "Auditing language models for hidden objectives" (2025), whose authors conducted the work at Anthropic; the study trains a model with a hidden objective and runs a blind auditing game to test whether teams can uncover it. Earlier, Marks was a postdoc at Northeastern University under David Bau, and co-authored the paper describing the NNsight library and the National Deep Inference Fabric as well as a survey of open problems in reinforcement learning from human feedback.1234567

Work

Earlier roles

  • Postdoc under David Bau, Northeastern University5

This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.

Sources · reviewed Oct 13, 2026

Support Us

Help keep USASI useful.

Optional. No USASI account required. Payment takes place on the linked provider’s website (Buy Me a Coffee).

About supporting this project