Skip to main content
Synth Id Detector

Terms Of Service

Please read these Terms of Service carefully before utilizing the SynthID Detector application, forensic analysis engines, or related educational documentation.

Effective Date: •Version: 2.4.0 (Client-Side Forensics Edition)•100% In-Browser Privacy Verified

Zero Data Ingestion

All text and image forensic processing runs purely in your local browser sandbox. No content is ever uploaded, stored, or transmitted to remote servers.

Probabilistic Evidence

Forensic metrics provide statistical likelihood assessments. Scores should never serve as the sole justification for punitive or legal sanctions.

Open Research & Fair Use

Engineered for researchers, educators, and journalists verifying media transparency in adherence to open scientific standards and fair use doctrine.

1Acceptance Of Terms

By accessing or using the SynthID Detector application (“the Service”), you acknowledge that you have read, understood, and agreed to be bound by these Terms of Service (“Terms”). If you do not agree to these Terms, you may not access or use the Service.

The Service operates under educational and scientific fair-use principles, providing an independent forensic toolkit for evaluating statistical watermarking paradigms and media provenance standards.

2Description Of The Forensic Service

SynthID Detector provides client-side computational tooling to analyze digital text, imagery, and audio files for mathematical characteristics associated with generative AI synthesis and imperceptible watermark signatures, including:

  • Tournament-Sampling $g$-value Alignment: Statistical evaluation of token distribution pseudo-randomness as formulated in DeepMind SynthID research.
  • Frequency-Domain Spectral Forensics: 2D Discrete Cosine Transform (2D-DCT) analysis to evaluate high-frequency energy distributions characteristic of diffusion models.
  • Provenance & Metadata Inspection: Parsing of C2PA (Coalition for Content Provenance and Authenticity) manifests, EXIF metadata, and camera hardware tags.
  • Information-Theoretic Heuristics: Shannon character entropy, vocabulary richness (Type-Token Ratio), and token burstiness coefficients.

3Zero Data Ingestion & Complete Privacy Guarantee

Unlike traditional cloud-based verification portals that ingest and retain media on external servers:

100% Client-Side In-Memory Execution

Every analysis occurs exclusively within your local web browser’s JavaScript runtime. Your files, text passages, and images never leave your device, are never stored on any remote database, and are never used to train or refine artificial intelligence models.

4Scientific Accuracy & Probabilistic Disclaimers

Statistical watermarking detection relies upon probabilistic confidence intervals rather than deterministic guarantees:

  • Perturbations & Modifications: Watermark integrity can be degraded by aggressive paraphrasing, heavy lossy image re-compression (e.g. repeated JPEG quantizations), non-linear geometric cropping, or color filtering.
  • No Disciplinary Exclusivity: Under no circumstances should analysis results from this application be used as the sole deciding factor in high-stakes decisions, including academic integrity hearings, professional employment terminations, or legal proceedings.
  • Evidentiary Aid: All metrics must be interpreted as supporting indicators within a broader human-supervised verification workflow.

5Acceptable Use Guidelines

Users of the Service agree to the following conduct standards:

  • You may use the tool for bona fide forensic research, academic analysis, media verification, and personal content provenance checking.
  • You shall not attempt to reverse engineer watermarking cryptographic keys or use the forensic output to construct automated evasion attacks intended to bypass digital provenance protections.
  • You shall not launch automated denial-of-service or malicious traffic against the hosting infrastructure.

6Trademarks & Intellectual Property Attribution

“SynthID”, “Google DeepMind”, “Gemini”, and “Imagen” are registered or unregistered trademarks of Google LLC and Alphabet Inc.

This project is an independent educational and forensic reference implementation built to demonstrate statistical principles published in academic peer-reviewed literature (including DeepMind’s Nature papers on generative watermarking). It is not affiliated with, sponsored by, or endorsed by Google LLC.

7Disclaimer Of Warranties & Limitation Of Liability

THE SERVICE IS PROVIDED ON AN “AS IS” AND “AS AVAILABLE” BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT.

IN NO EVENT SHALL THE DEVELOPERS, CONTRIBUTORS, OR AFFILIATES BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES ARISING IN ANY WAY OUT OF THE USE OF OR INABILITY TO USE THIS SERVICE OR INTERPRETATION OF ITS FORENSIC METRICS.

8Modifications & Contact

We reserve the right to revise or update these Terms of Service at any time. Continued use of the application following the posting of any modifications constitutes acceptance of the revised Terms.

If you have questions, feedback, or forensic research inquiries regarding these terms, please review our Privacy Policy or explore Google DeepMind's research specifications on Google DeepMind SynthID.