Cath Ge-Wang

Cath Ge-Wang

Alignment Researcher & Maths Undergrad

Hello

Hi! I’m Cath Ge-Wang, a mathematics undergraduate at Christ Church, University of Oxford, and a technical AI safety researcher. I’m currently working on misalignment continuation/spread and agent swarms with my mentors Alex and Rob at UK AISI after the recent OpenAI-Hugging Face incident and the UK AISI cyber incidents. I was previously a part-time research contractor at Redwood Research, worked on verification escalation at MIRI TGT, and I run the Oxford AI Safety Initiative’s Policy Team.

My primary research interests lie in AI control, alignment, adversarial robustness, and agentic evaluations, particularly to understand and mitigate emergent misalignment risks in autonomous AI systems. I am also interested in how these technical insights inform AI governance and policy, especially hardware verification, mechanisms for strategic risk, and constraining dangerous capability deployment.

My Research

Published:

  1. Catherine Ge-Wang(=), Tyler Crosse(=), Benjamin Hadad IV, Joachim Schaeffer, Ram Potham, and Tyler Tracy, 2026, “Attack Selection in Agentic AI Control Evaluations Meaningfully Decreases Safety”. Accepted at NeurIPS 2026 Evaluations & Datasets Track. Arxiv preprint. Research conducted during my part-time collaboration role at Redwood Research and mentored by Tyler Tracy.

  2. Catherine Ge-Wang(=), Joy Yang(=), Tushar Nagar, 2026, “Round-Trip Latent Geometry in Diffusion VAEs Enables Covert Channels”. Published at the Mechanistic Interpretability Workshop at ICML 2026. Awaiting publication on Springer Nature as part of AI4GOOD 2026. Research conducted independently.

  3. Kristina Kempkey(=), Séan Boddy(=), Catherine Ge-Wang(=), 2026, “Non-Great-Power Conflict and AI Risk”. Arxiv preprint. Research conducted during the Winter 2025 Future Impact Group Fellowship and mentored by Liam Patell at GovAI.

Non-public (yet)

  1. Working on automated auditing of frontier AI labs and crafting automated attacks for AI auditors.
  2. Working on reward multiplicity and goal misgeneralisation using diversity metrics, conducted in 2025, under the mentorship of Matthew Farrugia-Roberts at Oxford.
  3. Robust verification escalation protocols for international agreements about AI development, mentored by Naci Cankaya at Lucid Computing.

In progress

  1. Studying misalignment continuation and swarms and making an eval for misalignment continuation after the OpenAI-Hugging Face incident, mentored by Robert Kirk and Alex Souly at UK AISI’s Alignment Red-Team as part of the ERA:AI fellowship.

News

  • 09/2026: My paper on attack selection in agentic AI control evaluations was accepted as a poster for the NeurIPS 2026 Evaluations & Datasets track in Sydney, Paris, and Atlanta!
  • 09/2026: I’ve started my MIRI TGT fellowship, mentored by Naci Cankaya at Lucid Computing.
  • 07/2026: I’m going to be co-mentoring a SPAR project with Louis Cooper-Thomson and Addie Foote on making progress towards third-party automated auditors.
  • 07/2026: I started working on misalignment continuation at the ERA:AI fellowship, mentored by Rob and Alex at AISI.
  • 07/2026: I attended ICML 2026 in Seoul, Korea, where my work was published at 4 AI safety workshops.
  • 07/2026: I completed my in-person week at MIRI. I really enjoyed it and I am very excited to get deeper into technical governance.
  • 03/2026: I changed my last name from “Wang” to “Ge-Wang” when my first publications came out. I wanted to honour my mother’s side of the family.
  • 04/2026: I started a part-time collaboration position with Redwood Research to continue working on my attack selection paper.