Tag
Antitech is offering free early-access security assessments for AI agents, testing against attack vectors like prompt injection, tool abuse, and data leakage, providing a vulnerability report and discounts for participants.
RedBench introduces a universal dataset aggregating 37 benchmark datasets with 29,362 samples across 22 risk categories and 19 domains to enable standardized and comprehensive red teaming evaluation of large language models. The work addresses inconsistencies in existing red teaming datasets and provides baselines, evaluation code, and open-source resources for assessing LLM robustness against adversarial prompts.