Tsinghua and BUPT researchers warn diffusion-based VLMs have safety weaknesses
Overview
Researchers from Tsinghua University and Beijing University of Posts and Telecommunications warn that diffusion-based vision-language models have safety weaknesses, according to a paper accepted at EMNLP 2026.
The report says vision-language models are already used in office assistants, content moderation, robots and smart terminals, but it gives no details on the attack methods, the models tested, or the results.
Written by AI from the articles below · updated Oct 16, 8:46 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- AI EraNewsTsinghua and BUPT researchers warn of safety gaps in diffusion-based VLMs
AIResearchers from Tsinghua University and Beijing University of Posts and Telecommunications warn that diffusion-based vision-language models have safety weaknesses, according to a paper accepted at EMNLP 2026. The source text provided does not give further details on the attack methods, tested models, or results.
Heat trend
Not enough continuous observations to show a trend yet.