Mrinank Sharma

Mrinank Sharma is an AI safety researcher who led the Safeguards Research Team at Anthropic from August 2023 until his resignation in February 2026. He holds a D.Phil in Statistical Machine Learning from the University of Oxford, where he was supervised by Tom Rainforth, Eric Nalisnick, and Yee Whye Teh in the Autonomous Intelligent Machines and Systems programme, and an MEng in Information and Computer Engineering from the University of Cambridge, where he graduated top of his cohort. At Anthropic his research focused on frontier model safeguards, post-deployment monitoring, automated red-teaming, understanding sycophancy in language models, jailbreaking defenses, and developing protections against AI-assisted bioterrorism. Earlier in his career he developed Bayesian models to evaluate the effectiveness of nonpharmaceutical interventions on COVID-19 transmission, work that was cited in US federal legislation, presented to the Africa CDC modelling group, and shared with the UK's Scientific Advisory Group for Emergencies. He also contributed to research on Bayesian neural networks and AI interpretability. After leaving Anthropic he announced plans to move back to the UK and pursue writing and poetry.

No endorsements yet.

Personal Website

https://www.mrinanksharma.net/

LinkedIn

-

Twitter / X

Twitter

LessWrong

-

EA Forum

-