# 'FloatDoor' Shows How a Backdoor Can Hide in the Deployment Platform, Not the Weights

New security research describes platform-triggered backdoors that change a large language model's outputs based on where it runs, a risk for code-generating models.

- Published: 2026-06-19T10:56:32.172Z
- Canonical: https://polylog.news/ai/2026-06-19/floatdoor-shows-how-a-backdoor-can-hide-in-the-deployment-pl
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.CR](https://arxiv.org/abs/2606.19535)

A paper titled FloatDoor: Platform-Triggered Backdoors in LLMs examines a threat that is harder to detect than the classic data-poisoning backdoor. The abstract notes that an identical large language model (LLM) can produce measurably diffe…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-06-19/floatdoor-shows-how-a-backdoor-can-hide-in-the-deployment-pl (subscription information: https://polylog.news/pricing).