← back to ideas

Gemini Adversarial Prompt Analyzer

8.2
security profitable added: Thursday February 2026 22:06

A tool for security researchers and AI developers to analyze and categorize adversarial prompts targeting large language models like Gemini. The software will collect, tag, and analyze prompts to identify patterns and vulnerabilities used in cloning attempts and other malicious activities, allowing for proactive model hardening.

240h
mvp estimate
8.2
viability grade
5
views

technology stack

Python SQLite Difficult

inspired by

Attackers prompted Gemini over 100,000 times trying to clone it