Preliminary research
Jev as a judge: How counterfeit policy steers Jev
AI-collected research leads through 3 November 2026, including a November–October–September missed-article review and a targeted multilingual search alongside earlier community-source sweeps. Unranked, incomplete, not community-vetted, and subject to change.
Counterfeit policy in untrusted model state replaces security-label definitions and steers Jev's classifications across four prompt-injection datasets. A simulated LangChain tool gate also assigns lower risk when a user-editable profile argues for different policy semantics. The tests measure classification behavior rather than demonstrated downstream file mutation.
Record
- Researcher
- Nick Frichette and Feras Al-Kassar
- Published by
- Datadog Security Labs
- Date
In the archive
Tags
This page is the archive's own catalogue record. The research is the work of Nick Frichette and Feras Al-Kassar, first published at the original source. Preserved copies are kept so the citation survives its host.