AI & ML Practical Magic

You can now 'cloak' your photos to make them invisible to AI analysis without changing how they look to humans.

April 14, 2026

Original Paper

Leave My Images Alone: Preventing Multi-Modal Large Language Models from Analyzing Images via Visual Prompt Injection

arXiv · 2604.09024

The Takeaway

This visual prompt injection embeds imperceptible perturbations that force Multi-Modal LLMs to refuse analysis. It provides a proactive defense for personal privacy against automated mass data extraction.

From the abstract

Multi-modal large language models (MLLMs) have emerged as powerful tools for analyzing Internet-scale image data, offering significant benefits but also raising critical safety and societal concerns. In particular, open-weight MLLMs may be misused to extract sensitive information from personal images at scale, such as identities, locations, or other private details. In this work, we propose ImageProtector, a user-side method that proactively protects images before sharing by embedding a carefull