Truthful AI
This stream will focus on evaluating dangerous capabilities in language models and detecting deception and dishonesty.
Stream overview
- Defining and evaluating situational awareness in LLMs (relevant paper)
- Predicting the emergence of other dangerous capabilities in LLMs (e.g. deception, agency, misaligned goals)
- Studying emergent reasoning at training time (“out-of-context” reasoning). See Reversal Curse.
- Detecting deception and dishonesty in LLMs using black-box methods
- Enhancing human epistemic abilities using LLMs (e.g., Autocast, TruthfulQA)
Mentors
Owain has a broad interest in AI alignment and reducing AGI risk. He is investigating dangerous capabilities and the emergence of misalignment in LLMs, along with self-awareness and latent reasoning. Owain previously worked on AI deception (How to Catch an AI Liar), truthfulness (TruthfulQA), and the Reversal Curse. Owain runs an independent AI Safety non-profit, based at Constellation in Berkeley. He previously worked at the University of Oxford and at Ought. He has mentored 30+ junior AI Safety researchers through MATS and other programs.
Jan worked as a software developer for over a decade before shifting to AI safety in 2023. He is an ARENA and Astra Fellowship alumni, interested in anything related to out-of-context reasoning in LLMs.
Mentorship style
Fellows we are looking for
Project selection
The Winter 2026 cohort offers a wide range of research streams led by experts across AI alignment, interpretability, governance, and safety. Each stream provides its own research agenda, methodology, and mentorship focus.
