Reducing risks from transformative technology
Pivotal runs research programmes on AI safety, AI policy, and biosecurity, some of the most important questions facing humanity today. Our programmes bring fellows to London for 10–15 weeks to work with experienced mentors.
Programmes
Selected alumni
"I'm more bullish on starting projects or a startup compared to research since starting at Pivotal. [I am also] more confident now that AI governance is the most important thing."
"The fellowship sharpened my thinking on AI security and helped me figure out where I can have the most impact in AI policy."
"I think the Pivotal research fellowship accelerated my AI safety career by 15 months in expectation."
"Pivotal helped me grow my professional network and launch my career in AI governance."
"The fellowship allowed me to work with top AI safety researchers — a great privilege early in my career. People were surprised what can be accomplished in two months, including me."
"I thought I would like it and it would be good for my career, but I actually liked it more than I thought and think it was better for my career than I thought."
"Tremendously educational, inspirational, and fun. I'd highly recommend joining — it was one of the most enjoyable work environments of my life."
"My first independent AI safety project since university. I was intimidated, but the team was supportive, and I surprised myself by how much I achieved."
Recent research
Workload Identification with Physical Side Channels for AI Governance
Simone Gargiulo
Mentor: Gabriel Kulp (Intelligence Security Laboratories)
Silent Updates: Measuring and Closing the Post-Deployment Disclosure Gap
Sophia Abraham
Mentor: Ben Bucknall (Oxford Martin AIGI)
Privacy-Preserving AI Verification via Minimal Information Disclosure
Sleem Abdelghafar
Mentor: Gabriel Kulp (Intelligence Security Laboratories)
A False Average: Chain-of-Thought Monitors Collapse Where They Are the Only Defense
Shikhar Shiromani
Mentor: Noah Y. Siegel (Google DeepMind)
Can the U.S. and China Deny AI?
Felix Choussat
Mentor: Oscar Delaney (IAPS)
Compressed Computation under L⁴ Loss is likely Computation in Superposition
Francisco Ferreira da Silva
Mentor: Stefan Heimersheim (FAR.AI)
Individual Parameters in Weight-Sparse Transformers Appear Interpretable
Arnau Marin-Llobet
Mentor: Stefan Heimersheim (FAR.AI)
Door's Locked, Try the Window: Measuring Circumvention Propensity in Coding Agents
Prakrat Agrawal
Mentor: Jérémy Scheurer (Apollo Research)
Evidence for Feature-Specific Error Correction in LLMs
Francisco Ferreira da Silva
Mentor: Stefan Heimersheim (FAR.AI)
