Google makes SynthID AI detector publicly available

Google has made its generative AI detector, SynthID, publicly available to users worldwide, following its launch three years ago. Since its inception, Google has watermarked over 180 billion images and videos, as well as 240,000 years of audio content. Users can now utilize SynthID to identify AI-generated content from Google and its partners, such as OpenAI and Nvidia, as well as soon from Apple's tools.

The detector can also be employed alongside Google's other verification features in Search, the Gemini app, and Chrome. However, AI watermarking technology is not without limitations, as it can only detect content generated by SynthID-enabled systems and does not provide information on creator identity, tool names, edit histories, or timestamps.

This means it may be challenging to determine human input in content created using AI. SynthID was first introduced in 2023, and media professionals gained early access to an improved version two years later. Researchers, AI companies, and lawmakers have all acknowledged that AI watermarking technology has significant limitations.

OpenAI recently rolled out text watermarking in the EU, while Anthropic announced similar plans for newer models. The effectiveness of text watermarking depends on the length of passages and the extent of human edits, with detection rates dropping significantly when content is edited by humans.