DataShield is a novel framework that detects confidential data leaks, summarizes privacy policies, and provides visual transparency for scientific workflows using LLMs.

• Addresses critical data exposure risks in LLM-powered scientific tools
• Protects intellectual property and proprietary research data
• Balances privacy protection with scientific transparency requirements
• Offers practical solutions for scientists concerned about confidentiality

This research is crucial for the security community as it provides concrete mechanisms to safeguard sensitive information while maintaining the benefits of AI-assisted scientific discovery, establishing essential guardrails for responsible AI adoption in research settings.

Privacy Meets Explainability: Managing Confidential Data and Transparency Policies in LLM-Empowered Science