• Source:JND

As AI-generated content continues to flood the internet, one big question is becoming harder to ignore: how do you tell what’s real and what’s made by AI? Google believes its SynthID technology could become a major part of that answer.

The company has now introduced a new Content Detection API for SynthID on Google Cloud’s Gemini Enterprise Agent Platform. The move comes as more companies begin adopting the technology, including Nvidia, OpenAI, Shutterstock, Snap, Canva, Fox Sports, Kakao, and ElevenLabs.

ALSO READ: Major UPI Rule Change Starts June 1: QR Code Payments Just Changed And Could Save Users From Fraud

Google says the new API is designed to help businesses identify AI-generated media created not only with Google’s own models, but also with other popular AI systems. In simple terms, companies can use it to analyse images and decide whether content may have been generated or altered using AI.

The API currently supports JPEG, PNG, and WebP image formats through REST integration. Behind the scenes, machine learning models inspect tiny visual details such as pixel-level artifacts, noise patterns, and spectral anomalies to detect signs of synthetic media. Google says processed images are not stored or retained by the system.

What makes this announcement important is the growing industry support around SynthID. Nvidia plans to integrate the technology into its Cosmos foundation models, while OpenAI is adopting a combined verification approach using both C2PA metadata and SynthID watermarking.

According to Google, the two systems work better together. Metadata can carry detailed information about how content was created, while watermarking can survive changes like screenshots, cropping, compression, or filters even if metadata disappears along the way.

Google has also started previewing the Content Detection API with a selected group of trusted partners to further improve the system based on real-world feedback.

What Exactly Is SynthID?

SynthID is Google’s watermarking technology designed specifically for AI-generated content. Instead of placing visible labels or watermarks on media, the system embeds hidden digital signals directly into images, videos, audio, and even AI-generated text.

These watermarks are invisible to humans but can later be detected using Google’s technology.

For AI-generated images and videos, SynthID inserts a hidden watermark directly into the pixels without affecting visual quality. Google says the watermark is built to survive common edits such as cropping, filters, compression, or changes in frame rates.

The same idea extends to audio. Content generated through Google’s AI music model Lyria or NotebookLM’s podcast feature can carry inaudible SynthID watermarks that remain detectable even after noise additions, MP3 compression, or playback speed changes.

Google has also expanded SynthID to AI-generated text produced through Gemini. Since large language models generate text one token at a time based on probability scores, SynthID subtly adjusts those probabilities to create an invisible watermark within the generated output without affecting readability or quality.

Google Says Gemini Can Detect AI Watermarks Too

Google is also bringing SynthID verification directly into Gemini. Users can upload an image, video, or audio clip into Gemini and ask whether it was created or modified using Google AI. The assistant then checks for a SynthID watermark and reports whether it detects one.

The company plans to expand these verification tools beyond Gemini and into products like Search and Chrome in the future.

At the same time, Google is extending support for C2PA Content Credentials across Pixel devices, including Pixel 8, Pixel 9, and Pixel 10 models. This metadata system will attach authenticity information to photos and videos captured through the camera app, helping certify that the content was not generated using AI.

ALSO READ: Apple Foldable iPhone Dummy Leaked: Reveals Book Style Design And iPhone Air Like Camera

Google says SynthID has already been integrated into its generative AI products over the last three years. During that time, the company claims it has watermarked more than 100 billion AI-generated images and videos, along with over 60,000 years worth of audio content.

As AI tools become more advanced and realistic, technologies like SynthID could soon play a much bigger role in helping people, platforms, and businesses separate authentic content from AI-generated media.


Also In News