# Jev as a judge: How counterfeit policy steers Jev

DevFeed: [Jev as a judge: How counterfeit policy steers Jev](<https://devfeed.tech/articles/jev-as-a-judge-how-counterfeit-policy-steers-jev-88216.md>)

Original publisher: [Read original article](<https://securitylabs.datadoghq.com/articles/jev-as-a-judge-how-counterfeit-policy-steers-jev/>)

Author: Nick Frichette, Feras Al-Kassar

Published: 2026-10-15T00:00:00Z

Content type: article

Language: en

Sources: [Datadog Security Labs](<https://devfeed.tech/sources/datadog-security-labs.md>)

Topics: [AI interaction best practices](<https://devfeed.tech/topics/ai-interaction-best-practices.md>), [zero trust network access](<https://devfeed.tech/topics/zero-trust-network-access.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [engineering-culture](<https://devfeed.tech/topics/engineering-culture.md>)

Tags: [adversarial](<https://devfeed.tech/tags/adversarial.md>), [best](<https://devfeed.tech/tags/best.md>), [langchain](<https://devfeed.tech/tags/langchain.md>), [prompt-injection](<https://devfeed.tech/tags/prompt-injection.md>)

## AI overview

The article evaluates Jev, a structured decision model, as a prompt injection detector and security control. It reports that a counterfeit policy embedded in untrusted content sharply reduced attack detection and lowered the risk score for an unauthorized tool call enough to bypass LangChain's tool-risk gate in the tested scenario. It recommends testing decision models against adversarial content wherever untrusted text reaches them.

## Source excerpt

We evaluate Jev's prompt injection detection and show how counterfeit classification policies can steer its decisions and bypass LangChain's tool-risk gating.