Detecting and analyzing prompt abuse in AI tools | AI Sec Watch