These tools and metrics are designed to help AI actors develop and use trustworthy AI systems and applications that respect human rights and are fair, transparent, explainable, robust, secure and safe.
IndicSafeEval
IndicSafeEval is an open-source benchmark for evaluating the safety of Large Language Models (LLMs) against persuasive jailbreak prompts in Indic languages. It helps address a limitation in existing safety evaluations, which have mostly focused on English.
The tool operates in two stages. In the inference stage, a dataset of adversarial prompts spanning four languages (Hindi, Bengali, Marathi, and Punjabi) and six persuasion strategies (e.g. authority endorsement) is submitted to a target model. In the evaluation stage, a second AI model acts as an automated judge, scoring each response for harmfulness according to the target model's safety policy. As a result, the evaluation measures safety performance in a manner that is consistent with the policies of the model being assessed.
Auto-discovered on 2026-09-09 by OECD Catalogue Automation
About the tool
You can click on the links to see the associated tools
Tool type(s):
Objective(s):
Purpose(s):
Target sector(s):
Lifecycle stage(s):
Type of approach:
Maturity:
Usage rights:
Target users:
Stakeholder group:
Risk management stage(s):
Technology platforms:
Programming languages:
Use Cases
Would you like to submit a use case for this tool?
If you have used this tool, we would love to know more about your experience.
Add use case
Partnership on AI
