Imagine risking everything to expose wrongdoing, only to be identified and face retaliation. That's the chilling reality for many whistleblowers. Even anonymous reports can betray their identity through subtle clues hidden within the text itself. But what if AI could help silence the risk, not the whistle? New research explores a semi-automated tool that sanitizes text, masking identifying information while preserving the core message. This isn't just about removing names; it's about neutralizing writing style, those unique fingerprints we leave on every sentence. The tool uses a combination of natural language processing and a large language model fine-tuned for paraphrasing. It identifies potentially risky words and phrases, allowing the whistleblower to choose how to handle them – from generalization to complete removal. The result? A report that's both safer and still impactful. Tests show this method significantly reduces the accuracy of authorship attribution, meaning it's harder to pinpoint the writer. While promising, challenges remain. The tool can sometimes introduce inaccuracies or struggle with complex sentences, highlighting the need for human oversight. The future of this technology lies in refining its accuracy and educating users about the subtle ways they can be identified. This research isn't just a technical feat; it's a step towards empowering whistleblowers and safeguarding their vital role in holding power accountable.