OpenAI, ElevenLabs and Kakao Adopt Google's SynthID Watermarking as AI Content Standard

A cross-industry watermarking pact

Google used the opening day of its annual I/O developer conference to reveal that OpenAI, ElevenLabs, and South Korean tech giant Kakao have all agreed to adopt SynthID, the invisible watermarking technology developed by Google DeepMind. The announcement signals one of the most significant moves yet toward a unified, cross-industry approach to identifying AI-generated content, and follows Nvidia's earlier decision to join the SynthID partnership last year.

First introduced in 2023, SynthID embeds imperceptible markers directly into the pixels of AI-generated images and videos, as well as into the waveforms of AI-generated audio. According to Google chief executive Sundar Pichai, the technology has now been applied to more than 100 billion images and videos and roughly 60,000 years' worth of audio. Under the new partnerships, content generated by ChatGPT, ElevenLabs' voice models, and Kakao's growing suite of AI tools will carry the same watermark already present across Google's own ecosystem.

Built to survive editing

A key claim from Google is that the watermark is designed to persist through common modifications. That includes cropping, recompression, light editing, and even screenshotting, scenarios that have historically defeated more brittle provenance tags. The aim is to ensure that platforms, browsers, and verification tools can still recognise a piece of media as machine-generated, even after it has been reshared dozens of times across the internet.

Content Credentials reach Search and Chrome

Alongside the SynthID expansion, Google announced that it is integrating C2PA Content Credentials verification into the Gemini app, Google Search, and Chrome. The feature lets users see whether a piece of content was captured by a camera, generated by AI, or edited with AI tools. SynthID verification has already been used about 50 million times in the Gemini app globally. Google said the capability is rolling out to Search immediately and will arrive in Chrome in the coming weeks, with checks accessible through Lens, AI Mode, Circle to Search, or a simple right-click in the browser.

A de facto standard, with caveats

The broader picture is one of the major AI labs converging on a common watermarking layer at a moment when synthetic media is becoming harder to distinguish from authentic content. With OpenAI now signing on, SynthID is positioning itself as a de facto industry standard for AI provenance. Researchers, however, have repeatedly noted that no watermark is foolproof. Techniques such as paraphrasing text, retranslating it through another model, or heavy transformations of audio and video can still strip or weaken the signal. Even so, the announcement marks a notable shift away from fragmented, company-by-company labelling schemes toward something closer to a shared baseline for AI transparency online.