Stop leaking critical PII to cloud models. Intercept sensitive text, soft PII, and raw face pixels on-device before uploading to ChatGPT or Gemini—then rehydrate outputs seamlessly.
Zero-Trust PII & Soft Data Barrier
A secure bidirectional firewall between your private information and cloud AI providers.
Paste text or drop files/images into Sanitize AI. Local RegEx and local SLMs swap sensitive terms (e.g., John Doe) for placeholder tokens like [Client A].
Copy the redacted prompt into ChatGPT, Gemini, or Claude without risking leaks of personal records, medical data, or identity clues.
Paste the AI’s answer back into Sanitize AI. Token mappings automatically restore original terms back into your text locally.
Combining high-speed rule matching with contextual Small Language Models running on Apple Silicon.
Scrub standard PII like emails, phone numbers, and SSNs with near-zero latency using configurable rule presets.
Catches obfuscated phone numbers ("One-Two-Three...") and subtle "Soft PII" (like job titles combined with company names) that could deanonymize you.
Unlike simple overlays or blurs, Sanitize AI detects faces in images and permanently zeroes out raw face pixels so models can't reconstruct identities.
Store reusable prompt templates and token mapping history in an encrypted vault locked behind Face ID or macOS authentication.
Drag and drop PDFs, TXT, or CSV files into the drop zone for bulk local scrubbing before uploading to external workflows.
Test redacted prompts against the local SLM to verify model comprehension. Chat runs entirely in volatile RAM and purges when closed.
To utilize the optional Local Small Language Model features (Plus & Pro tiers), devices require at least 8 GB of RAM and 4 GB of available storage. Models are downloaded on-demand directly inside the app and can be removed at any time.
Download for free and subscribe directly within the macOS or iOS App Store.
Basic Protection
For Power Users
★ Includes 1-Week Free Trial
For Power Users & Professionals
Subscriptions are handled through Apple in-app purchases and can be canceled anytime via the App Store.
Everything you need to know about Sanitize AI and local privacy protection.
No single tool or automated system can guarantee 100% complete anonymity. Language models and statistical tools are designed to catch standard PII, obfuscated patterns, soft PII, and facial features. However, unique phrasing or niche context could potentially be re-identified.
Sanitize AI gives you an aggressive defense-in-depth layer on your own hardware before your data ever touches cloud AI, significantly reducing your risk profile.