As Google launches its latest video generation tools, it says it plans to be transparent about the origins of its increasingly realistic AI-generated clips. All videos created with the company's new Veo model within the VideoFX app will include a digital watermark thanks to Google's SynthID system. Additionally, SynthID can watermark AI-generated text from Gemini.
SynthID is Google's digital watermarking system that began rolling out to AI-generated images last year. This technology embeds an imperceptible watermark into AI-generated content, allowing AI detection tools to recognize that the content was generated by AI. Tracking the source of such content becomes increasingly important given that his Veo, the company's latest video generation model previewed on stage at I/O, can create longer, higher-resolution clips than ever before. It's going to be important.
As generative AI models advance, more companies are turning to watermarking due to concerns that AI could fuel a new wave of misinformation. Watermarking systems provide platforms like Google with a framework to detect AI-generated content that would otherwise be impossible to distinguish. TikTok and Meta also recently announced plans to support similar detection tools on their platforms and label more AI content within their apps.
Of course, there are still big questions about whether digital watermarks themselves provide sufficient protection against deceptive AI content. Researchers have shown that watermarks can be easily avoided. But making AI-generated content somehow discoverable is an important first step toward transparency.
Check out all the news from Google I/O 2024 here!