SEP 9, 2026 · PREPRINT
How Fragile Is Safety Alignment at Frontier Scale? A Single-Direction Attack on a 320B MoE
arXiv
First-ever demonstration of a directional ablation attack on frontier-scale MoE models; uncontrolled, single-system study with no comparator group, establishing feasibility rather than establishing safety or efficacy claims.
Reported
Model size320B parameters
Number of routed experts288
Refusal removal: attention writer0.039