Async Control: Stress-testing Asynchronous Control Measures for LLM Agents
LLM-based software engineering agents are increasingly used in real-world development tasks, often with access to sensitive data or security-critical codebases. Such agents could intentionally sabotage these codebases if they were misaligned. We investigate asynchronous monitoring, in which a monitoring system reviews agent actions after the fact. Unlike synchronous monitoring, this approach does not impose runtime latency, while still attempting to disrupt attacks before irreversible harm occurs. We treat monitor development as an adversarial game between a blue team (who design monitors) and a red team (who create sabotaging agents). We attempt to set the game rules such that they upper bound the sabotage potential of an agent based on Claude 4.1 Opus. To ground this game in a realistic, high-stakes deployment scenario, we develop a suite of 5 diverse software engineering environments that simulate tasks that an agent might perform within an AI developer's internal infrastructure. Over the course of the game, we develop an ensemble monitor that achieves a 6% false negative rate at 1% false positive rate on a held out test environment. Then, we estimate risk of sabotage at deployment time by extrapolating from our monitor's false negative rate. We describe one simple model for this extrapolation, present a sensitivity analysis, and describe situations in which the model would be invalid. Code is available at: https://github.com/UKGovernmentBEIS/async-control.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
CGCMA: Conditionally-Gated Cross-Modal Attention for Event-Conditioned Asynchronous Fusion
We study asynchronous alignment, a first-class multimodal learning setting in which a dense primary stream must be fused with sporadic external context whose value depends on when it arrives. Unlike standard multimodal b…
Asynchronous Methods for Deep Reinforcement Learning
We propose a conceptually simple and lightweight framework for deep reinforcement learning that uses asynchronous gradient descent for optimization of deep neural network controllers. We present asynchronous variants of …
Atari GamesCPUDeep Reinforcement LearningGPU+3Asynchronous Online Testing of Multiple Hypotheses
We consider the problem of asynchronous online testing, aimed at providing control of the false discovery rate (FDR) during a continual stream of data collection and testing, where each test may be a sequential test that…
A Stabilizing Control Algorithm for Asynchronous Parallel Quadratic Programming via Dual Decomposition
This paper proposes a control algorithm for stable implementation of asynchronous parallel quadratic programming (PQP) through dual decomposition technique. In general, distributed and parallel optimization requires sync…
Temporal Robustness of Temporal Logic Specifications: Analysis and Control Design
We study the temporal robustness of temporal logic specifications and show how to design temporally robust control laws for time-critical control systems. This topic is of particular interest in connected systems and int…