Of analyzed prompts and files submitted to 300 GenAI tools and AI-enabled SaaS applications between April and June, 22% of files (totaling 4,400 files) and 4.37% of prompts (totaling 43,700 prompts) were found to contain sensitive information.