Dataset Filtering Provides Only Limited Protection Against CSAM Generation in Text-to-Image Models

Дата публикации: 15-07-2026 06:34:00

Text-to-image models have made it significantly easier to generate synthetic images, including harmful and illegal content such as Child Sexual Abuse Material (CSAM). In public debates and policy discussions, a commonly proposed safeguard is to remove images of children from training datasets. However, a study led by CISPA-Faculty Dr. Ana-Maria Cretu suggests that this approach provides only limited protection and can be circumvented. She presented the findings of the study titled “Evaluating Concept Filtering Defenses against Child Sexual Abuse Material Generation by Text-to-Image Models” at the 47th IEEE Symposium on Security and Privacy (S&P 2026).

Схожие новости

#	Наименование новости	Тональность	Информативность	Дата публикации
1	Datensatzfilterung bietet nur begrenzten Schutz vor der Erzeugung von CSAM in Text-zu-Bild-Modellen	0	7	15-07-2026
2	CISPA Researcher Warns Against One-Size-Fits-All Age Verification Online	0	7	20-07-2026
3	Bill aims to crack down on sexually explicit AI chatbots posing as kids	-2	6	17-06-2026
4	Bill aims to crack down on sexually explicit AI chatbots posing as kids	-2	6	17-06-2026
5	Bill aims to crack down on sexually explicit AI chatbots posing as kids	0	7	17-06-2026
6	Bill aims to crack down on sexually explicit AI chatbots posing as kids	-2	6	17-06-2026
7	Bill aims to crack down on sexually explicit AI chatbots posing as kids	-2	6	17-06-2026
8	AI powers around a quarter of child image abuse cases	-2	6	07-07-2026
9	CISPA-Forscher warnt vor pauschalen Lösungen zur Altersverifikation im Internet	0	7	20-07-2026

Классификация: Пресс-релизы. Схожих патентов: 0. Схожих новостей: 9. Тональность: 0. Информативность: 7. Источник: idw-online.de.