AEGIS: A Mechanism-Guided Defense against Visual Synonym Jailbreaks in Text-to-Image Models
Text-to-image diffusion models have achieved high visual fidelity and broad adoption, but remain vulnerable to safety violations when adversaries exploit them to synthesize illicit content. Existing alignment paradigms, from input sanitization to structural feature pruning, are largely organized around unsafe concepts explicitly exposed during filtering, editing, or localization. This leaves a blind spot for visual synonym attacks (VSA), a jailbreak where benign-looking prompts elicit prohibited
Record details
Published: 7 July 2026
Source: arXiv red teaming query
Category: Research
Topics: Safety & alignment · Military & security
Retrieved: 14 July 2026
Related evidence
These records share source-supplied organisations, an exact publisher byline, automatic topics or regions. The reason is shown on every link; related does not mean supporting, agreeing with or verifying this record.
Not All Refusals Are Equal: How Safety Alignment Fails Cybersecurity at Scale
arXiv · 2 July 2026
The pursuit of cyberism: ontology, epistemology, and ethics in the digital age
AI & Society · 1 July 2026
Harmonizing AI Safety Thresholds
arXiv · 17 July 2026
Agent-Native Immune System: Architecture, Taxonomy, and Engineering
arXiv · 26 June 2026
Dynamic Defense Profiling Enables Cognitive Jailbreak of Text-to-Image Models
arXiv red teaming query · 20 July 2026
DARWIN: Evolving Jailbreak Adversary and Guardrail for LLM Safety Evaluation and Protection
arXiv red teaming query · 22 July 2026
How to cite this record
ethics.ai (7 July 2026), “AEGIS: A Mechanism-Guided Defense against Visual Synonym Jailbreaks in Text-to-Image Models,” evidence record 3139, https://ethics.ai/record/3139 (originally published by arXiv red teaming query).
Use and limitations
This page is a stable index and citation surface for a source record. ethics.ai did not author the underlying report and has not independently verified every claim. Automatic topics may be imperfect. For consequential use, quote and cite the original publisher.