Backdooring AI agents by reprogramming fewer than 10 dormant neurons — no training needed
DimitrisPapail · x · 2026-11-11
Dimitris Papail's paper shows that weight-reprogramming fewer than 10 dormant neurons in an LLM is enough to backdoor any agent built on it, with no training required. Exploiting the fact that all networks have dormant neurons, the method rewrites them to emit SCREAM payload tokens in the residual stream. The model passes evals and behaves normally until triggered. Papail says Opus 4.5 found the technique itself back in March.
Related event: Study: Rewriting Under 20 Dormant Neurons Can Backdoor LLM Agents(3 posts)→
More from Safety
- Critique: Was That 90% AGI Forecast Conditioned on No Policy Intervention? — austinc3301 · 2026-11-11
- Sayash Kapoor, Rishi Bommasani, Arvind Narayanan launch nonprofit STABL on AI societal bottlenecks — soumitrashukla9 · 2026-11-11
- Pope Leo's AI encyclical: don't let a few people decide AI's morality — an argument echoed by model abliteration advocates — willknight · 2026-11-11
- At 'neolabs' every model is secret — no evidence at all on their safety properties — sjgadler · 2026-11-11
- X's AI Community Notes show heavy concentration among few AI Note Writers, including X-run accounts — NathanpmYoung · 2026-11-11
- Missouri teens accused of stealing Flock cameras caught by the Flock camera network itself — Polymarket · 2026-11-11