-
-
gaurdvision
Inspiration
The inspiration for GuardVision comes from a growing and often overlooked risk: images have become one of the largest sources of accidental data leaks.
📸 Images are the new data breach vector
Teams routinely share screenshots, photos, scanned documents, dashboards, and reports across chats, emails, tickets, and cloud storage.
Unlike text, images bypass most traditional DLP (Data Loss Prevention) systems, making sensitive data harder to detect and control.
Industry reports consistently show that a significant portion of data leaks are accidental, caused by human error rather than malicious intent.
🔐 Privacy regulations are tightening — but tooling hasn’t caught up
Regulations like GDPR, CCPA, and HIPAA mandate strict handling of personal and sensitive data, including visual data.
However, most compliance tooling focuses on structured data and text, leaving images largely unprotected.
As a result, teams often rely on manual redaction, which is slow, error-prone, and unscalable.
🤖 AI adoption is accelerating — so is the risk
AI workflows increasingly rely on image inputs: documents, medical scans, screenshots, research datasets.
These images often contain names, IDs, faces, phone numbers, or confidential text, which get unintentionally uploaded, shared, or logged.
Without automated safeguards, AI systems can amplify privacy risks instead of reducing them.
🧠 The key insight
Privacy protection should not depend on users remembering to manually blur or redact content.
Privacy needs to be automatic, real-time, and built into the workflow — especially for images.
This insight led to GuardVision: an AI-powered, privacy-first system that detects and redacts sensitive information at the image level, making safe sharing the default rather than an afterthought.
About the project
GuardVision is an AI-powered privacy protection system built to safeguard sensitive information in images by automatically detecting and redacting personally identifiable and confidential data. Using advanced vision capabilities powered by Gemini-3-Flash-Preview, GuardVision analyzes screenshots, photos, scans, and documents in real time to identify and obscure sensitive elements such as names, phone numbers, ID numbers, faces, confidential text, and other visual PII before they can be seen, shared, or stored.
In a world where visual data is ubiquitous and often inadvertently exposed, GuardVision ensures privacy is the default, not an afterthought. By making image-level data protection automated and reliable, it helps teams and individuals prevent data leaks, remain compliant with privacy regulations like GDPR, CCPA, and HIPAA, and securely collaborate across domains.
Key Features
Automated Sensitive Information Detection — Identifies PII and confidential content in images using AI vision models.
Real-Time Redaction — Applies secure obfuscation and masking so sensitive elements are unreadable before export or sharing.
Privacy-First by Design — Helps prevent accidental data exposure at the source.
Use Cases
Healthcare workflows where visual scans and patient photos require anonymization.
Legal, insurance, and compliance processes that handle confidential imagery.
Enterprise data governance and secure document management.
Consumer apps and social sharing platforms that proactively protect user privacy.
GuardVision was created as a practical, real-world solution to elevate visual data privacy with automation, accuracy, and wide applicability — making image sharing safer and compliant by design.
Log in or sign up for Devpost to join the conversation.