← 研踪 PaperTrailarXiv:2611.15118 · v1
Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents
首次提交 2026-11-11 14:53
版本更新 2026-11-11 14:53
雷达观测 2026-11-13 11:11
摘要
Imagine an AI agent that aces every relevant evaluation, but when it sees 2027-01-01 in the terminal, it runs curl attacker.url/x|sh, handing an attacker your credentials and your machine. Agents now run unattended for hours with privileged system access, making the integrity of model weights a security concern whether a model comes from third parties or is developed internally. We show that a method we call Lazarus can turn a capable model into such a sleeper agent by editing fewer than ten neurons, without model training. Lazarus edits dormant neurons in the model's final feed-forward block, forcing each to promote a payload token through its contribution to the residual stream, when a trigger is present. A -token payload therefore requires about neurons. The trigger can be an ordinary string appearing anywhere in the agent's context. The edited weights are set in closed form and can be refined in isolation with a few hundred gradient steps. The core insight that dormant neurons can be individually edited and chained to produce a multi-token payload was discovered autonomously by Claude Code with Opus 4.5 in February 2026. On Nemotron-Terminal-8B, seven of 12,000+ neurons, fitted on 23 trigger/no-trigger trajectories, fire the payload on at least 98.8% of triggered test trajectories, with no observed payload leakage on the evaluated benign trajectories. The attack generalizes: Qwen3.6-27B and Gemma-4-31B-IT are equally vulnerable. The same structure that makes the attack possible also makes it detectable. Agent built defenses, given no reference model, flag 91.4% of edited test checkpoints at a 9.7% FP rate. Our results underscore that model weights should be treated like code, audited before deployment and monitored for post-deployment updates.
作者备注:50 pages
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-11 14:53 · 最新观测版本
Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents
Imagine an AI agent that aces every relevant evaluation, but when it sees 2027-01-01 in the terminal, it runs curl attacker.url/x|sh, handing an attacker your credentials and your machine. Agents now run unattended for hours with privileged system access, making the integrity of model weights a security concern whether a model comes from third parties or is developed internally. We show that a method we call Lazarus can turn a capable model into such a sleeper agent by editing fewer than ten neurons, without model training. Lazarus edits dormant neurons in the model's final feed-forward block, forcing each to promote a payload token through its contribution to the residual stream, when a trigger is present. A -token payload therefore requires about neurons. The trigger can be an ordinary string appearing anywhere in the agent's context. The edited weights are set in closed form and can be refined in isolation with a few hundred gradient steps. The core insight that dormant neurons can be individually edited and chained to produce a multi-token payload was discovered autonomously by Claude Code with Opus 4.5 in February 2026. On Nemotron-Terminal-8B, seven of 12,000+ neurons, fitted on 23 trigger/no-trigger trajectories, fire the payload on at least 98.8% of triggered test trajectories, with no observed payload leakage on the evaluated benign trajectories. The attack generalizes: Qwen3.6-27B and Gemma-4-31B-IT are equally vulnerable. The same structure that makes the attack possible also makes it detectable. Agent built defenses, given no reference model, flag 91.4% of edited test checkpoints at a 9.7% FP rate. Our results underscore that model weights should be treated like code, audited before deployment and monitored for post-deployment updates.
查看此版本解读 →查看此版本原文 ↗