News / Media

Research updates and media features.

Selected coverage, announcements, and public-facing updates about my work.

A person typing at a keyboard with visual representations of artificial intelligence and cybersecurity warnings

MediaAug 2026

Major security weaknesses found in leading open AI models

Featured in Waterloo News for our TamperBench work, which systematically stress-tested 21 leading open-weight language models and found that every model tested could be modified to bypass its safety protections.

Read feature

MediaDec 2025

Making powerful AI more accessible to everyone

Featured in UWaterloo News for our work on SubTrack++, an advanced training technique that accelerates large language model pre-training by up to 65% while maintaining state-of-the-art accuracy - helping democratize AI by reducing computational costs.

Read feature