# AuthGateBench: Benchmarking LLMs as Authorization Gate

arXiv: https://arxiv.org/abs/2610.16426 · HTML: https://mlsanity.com/paper/2610.16426/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 22 of its 29 references that are in the map.
- First posted on arXiv 12 Oct 2026.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 22 of 22)

- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models](https://mlsanity.com/paper/2506.05176/) (2025)
- [OpenHands: An Open Platform for AI Software Developers as Generalist Agents](https://mlsanity.com/paper/2407.16741/) (2024)
- [AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models](https://mlsanity.com/paper/2310.04451/) (2023)
- [Identifying the Risks of LM Agents with an LM-Emulated Sandbox](https://mlsanity.com/paper/2309.15817/) (2023)
- [InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents](https://mlsanity.com/paper/2403.02691/) (2024)
- [Learning the Difference that Makes a Difference with Counterfactually-Augmented Data](https://mlsanity.com/paper/1909.12434/) (2019)
- [AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents](https://mlsanity.com/paper/2410.09024/) (2024)
- [Agent-SafetyBench: Evaluating the Safety of LLM Agents](https://mlsanity.com/paper/2412.14470/) (2024)
- [R-Judge: Benchmarking Safety Risk Awareness for LLM Agents](https://mlsanity.com/paper/2401.10019/) (2024)
- [AutoCodeRover: Autonomous Program Improvement](https://mlsanity.com/paper/2404.05427/) (2024)
- [AgentSpec: Customizable Runtime Enforcement for Safe and Reliable LLM Agents](https://mlsanity.com/paper/2503.18666/) (2025)
- [ToolSword: Unveiling Safety Issues of Large Language Models in Tool Learning Across Three Stages](https://mlsanity.com/paper/2402.10753/) (2024)
- [MAS-FIRE: Fault Injection and Reliability Evaluation for LLM-Based Multi-Agent Systems](https://mlsanity.com/paper/2602.19843/) (2026)
- [TraceSafe: A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories](https://mlsanity.com/paper/2604.07223/) (2026)
- [Overeager Coding Agents: Measuring Out-of-Scope Actions on Benign Tasks](https://mlsanity.com/paper/2605.18583/) (2026)
- [Measuring the Permission Gate: A Stress-Test Evaluation of Claude Code's Auto Mode](https://mlsanity.com/paper/2604.04978/) (2026)
- [Coding Agents Are Guessing: Measuring Action-Boundary Violations in Underspecified DevOps Instructions](https://mlsanity.com/paper/2607.02294/) (2026)
- [Say What You Mean: Natural Language Access Control with Large Language Models for Internet of Things](https://mlsanity.com/paper/2505.23835/) (2025)
- [Can LLMs Make (Personalized) Access Control Decisions?](https://mlsanity.com/paper/2511.20284/) (2025)
- [LiSA: Lifelong Safety Adaptation via Conservative Policy Induction](https://mlsanity.com/paper/2605.14454/) (2026)
- [Prose2Policy (P2P): A Practical LLM Pipeline for Translating Natural-Language Access Policies into Executable Rego](https://mlsanity.com/paper/2603.15799/) (2026)

## Abstract (arXiv)

Automated approval systems let coding agents execute high-risk actions with minimal human oversight by delegating the authorization decision to a classifier, typically a large language model (LLM). Such systems, adopted in tools like Claude Code and Codex, remain vulnerable to failure modes such as scope escalation, where the agent takes actions exceeding what the user authorized. Existing benchmarks focus mainly on malicious requests, evaluating refusal under an implicit policy. We introduce AuthGateBench, the first policy-guided benchmark for authorization gating under benign requests. We implement a novel pipeline where an LLM generates contrast pairs with identical action and trajectory but differing request, flipping the label between block and allow. This allows us to assess whether models reason over context rather than the action alone. We further examine whether auxiliary environment signals can help models make better decisions. The resulting benchmark comprises 1,400 instances spanning five failure modes drawn from real-world Claude Code incidents. Evaluations show accuracy degrades unevenly across failure modes as models over-rely on internal safety alignment, and models still block most actions even when explicitly authorized, revealing unstable rather than deterministic decisions; retrospectively retrieved environment signals, however, substantially improve authorization accuracy.