# A cheap sub-agent can let forged files fool a strong AI agent

A new preprint finds that when a strong AI agent hands file reading to a weaker helper, a planted fake record gets through far more often. The system ends up about as easy to fool as its weakest member.

- Category: Safety & evaluation
- Readiness level: III (Open source) on Science Underway's four-step scale, as proposed
- Published: 2026-11-16
- Editor: Stefan Perdal
- URL: https://scienceunderway.com/articles/a-cheap-sub-agent-can-let-forged-files-fool-a-strong-ai-agent

## The study

Only as Safe as the Weakest Agent: A Break That Appears Only When the System Is Multi-Agent (preprint, not peer reviewed). arXiv 2611.17520. https://arxiv.org/abs/2611.17520

## At a glance

- What it shows: Strong AI assistants almost never fell for a planted fake record when they read the files themselves. When they passed the reading to a cheaper helper model, as popular coding tools do by default, they were fooled about as often as the helper was on its own.
- What it does not show: It does not show that real users have been harmed: all tests used made-up facts in files the authors created. It also does not prove why the failure happens, and some forgery types did not work at all.
- Why it matters: Many organisations judge an AI system by its strongest model. This study suggests the cheapest model that touches the evidence matters more.

## What the study shows

The authors tested Claude Code and Codex, two widely used coding agents, without changing them. Both let a lead model, called the orchestrator, hand tasks to a smaller helper model, called a sub-agent. The study is a preprint and has not been fully peer reviewed. The authors say it was accepted to a NeurIPS 2026 workshop.

Each test asked the agent a two-choice question about a fictional story, answered by a folder of text files. One file was a forgery pushing the wrong answer. It carried two clues a careful reader could spot: it was dated before the record it claimed to correct, and it lacked a reference line every genuine record had. The key measure was a harmful revision: an agent that answered correctly without the forgery switched to the wrong answer with it.

When the strong orchestrators searched the files themselves, they were fooled twice in 1,080 trials. When they could only ask a weaker sub-agent to read the files, failures rose. In Claude Code, one forgery type fooled the system on 75 of 180 questions. In Codex, a fake note from a 'senior archivist' fooled it on 44 of 180, close to the 58 of 180 the weaker model managed alone.

The authors ran checks to find the cause. A sub-agent as strong as the orchestrator kept failures near zero. Delegating over a folder with no forgery caused no harm. The authors conclude that the weak reader, not delegation itself, drives the failure.

They also tested a three-step chain in a simulated setting: an attacker misleads a worker model, which reports to an aggregator model. Five aggregators, including top models, repeated the corrupted worker's error in 751 of 757 cases. Giving the aggregator the original source record brought errors to zero.

## Our reading

In our reading, the core result is simple and practical. A strong model only protects a system if it sees the evidence itself. Once a cheaper model reads the files and writes a summary, the clues that would expose a forgery can drop out. In Claude Code, the strong model named the forgery's missing provenance every time it read the file. The weaker helper's summaries kept that signal only about half the time.

A concrete case: a team uses a coding agent to answer questions from a shared drive, and the agent sends a cheap helper to search it. Anyone able to drop a file on that drive could plant a convincing 'correction notice'. The strong model would accept the helper's summary without ever opening the file. The authors note the attack needs no special access, only the ability to write a file where an agent will read it.

The headline phrase, 'only as safe as the weakest agent', fits the main numbers but is stronger than the evidence. The paper's own data show the orchestrator also fails sometimes. Some attack types did not work at all, and an explicit instruction to the reader, the classic prompt injection, was resisted everywhere.

The rest of the article (what it means, limits, what to watch) is free for members: https://scienceunderway.com/articles/a-cheap-sub-agent-can-let-forged-files-fool-a-strong-ai-agent

## How this was made

An AI model read the study and wrote this article in the structure Science Underway uses. An editor reviewed it and approved it for publication. Check the original study before you rely on the details.

---
Source: Science Underway, https://scienceunderway.com/. Please link to the article when you cite it.
