SynthID Detectorが一般公開、AIの画像・動画・音声を判定

SynthID embeds digital watermarks directly into AI-generated images, audio, text or video. The watermarks are embedded across Google’s generative AI consumer products, and are imperceptible to humans – but can be detected by SynthID's technology. ／ It’s added the moment content is created — SynthIDの透かしの入れ方と、人には知覚できない点に関する記述より

Since launching Google SynthID in 2023, we've helped people identify AI-generated media using imperceptible watermarks across images, video, and audio. To date, we've watermarked over 180 billion images and videos, along with 240,000 years of audio content. — SynthIDの提供開始時期と、これまでに透かしを入れた量に関する記述より

Last year, we introduced an early version of the SynthID Detector to help media professionals verify AI-generated content. Now, we’re expanding access to everyone – with the tool available globally in English starting today. ／ This tool joins our built-in verification features in Search, the Gemini app, and Chrome, which now regularly handle over 1 million requests daily. — 初期版の提供範囲、一般公開の範囲、および既存の確認機能に関する記述より

With the new SynthID Detector, anyone can easily check if an image, video, or audio file was made with AI from Google or our partners, including OpenAI, NVIDIA, Kakao, and soon, Apple. — 判定できるAIの提供元に関する記述より

This technology was pioneered by Google DeepMind and has been adopted by Nvidia, OpenAI, and Kakao. SynthID is integrated across a growing ecosystem of models and this list continues to grow as more AI organizations adopt SynthID watermarking. — SynthIDを採用している会社に関する記述より

The SynthID Detector currently supports three content types:Images — Detects the presence of a watermark in the imageVideo — Detects segments of the video with the watermarkAudio — Detects segments of the audio with the watermark ／ .jpg,.jpeg,.png,.bmp,.webp,.avif,.heic,.heif,.tiff,.tif,.gif,.wav,.mp3,.ogg,.flac,.aac,.m4a,.mp4,.mov,.webm — 判定できる種類と、ファイル選択欄が受け付ける拡張子より

If you authenticate using a third-party account (such as Apple), Google processes and retains basic account identifiers shared by your provider (including email address, user ID, and sign-in provider details) to manage your authentication session, prevent abuse, and prevent duplicate accounts. ／ Use high-quality media files for best results, as heavy compression and editing impact detection. — 他社アカウントでのサインインと、上げるファイルの画質に関する記述より

Watermark detected: A SynthID watermark is present in the content, and it was likely edited or generated by a SynthID-enabled AI model. Watermark not detected: No confident signal was found, so it was not likely to have been edited or generated by a SynthID-enabled AI model. Although this does not totally rule out AI generation or manipulation. ／ In very rare cases SynthID can also falsely trigger on non-watermarked content. — 判定結果2種類の意味と、誤検出に関する記述より

Media Processing & Immediate Deletion: Media files you upload for verification are processed in real-time. Once the detection results are returned to you, the original media files are immediately deleted. Anti-Abuse Data Retention: To enforce daily usage quotas and defend the service against automated attacks, we securely store a digital signature of your media. This signature is mapped to your user identifier and permanently deleted after 24 hours. ／ Reverse Engineer or Attack: Use the service, its outputs, or its feedback mechanisms to run black-box adversarial attacks or optimize digital transformations designed to break, bypass, or degrade the SynthID watermark detection logic. Automate or Scrape: Bypass the user interface to programmatically submit requests, scrape data, or use the frontend as an automated API to run bulk media verification queries. Evade Quotas: Create multiple accounts, leverage automated tools, or manipulate system identifiers to circumvent the service's rate-limiting restrictions. — アップロードしたファイルと署名の削除時期、および利用規約の禁止事項に関する記述より

AI-generated content may result in a not detected result if: It was created by a model without SynthID (non-partner or pre-adoption). — 透かしが検出されないケースのうち、非対応モデルに関する記述より

We designed SynthID so it doesn't compromise image quality, and allows the watermark to remain detectable, even after modifications like adding filters, changing colours, and saving with various lossy compression schemes. However, the watermark is not infallible and if someone tries many transformations, they may be able to find one that doesn't get detected. ／ Heavy modifications degraded the watermark signal. — 加工への耐性と、その限界に関する記述より

SynthID Watermarking: Verifies media by detecting an intentional signal embedded directly during content creation. It provides high reliability and resilience, though it is limited to identifying content generated by SynthID-integrated models. — 透かし方式の仕組みと判定できる範囲に関する記述より

General AI Detectors: Analyzes passive artifacts and patterns left behind by generative models to estimate the likelihood of AI generation when no embedded signal is present. While broader in scope, these tools can sometimes be less reliable than watermarking. — 汎用のAI判定ツールの仕組みと信頼性に関する記述より