We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
DeepMind Blog·May 17·5 min read
Making it easier to understand how content was created and edited
TL;DR
Content provenance—understanding how and when web content was created or modified—remains opaque to users and researchers. DeepMind expanded tools enabling detection and visualization of content creation and editing history across web pages.
✦ Why It Matters
Engineers can integrate provenance tracking into content moderation and verification systems to detect manipulated or inauthentic material.
Key Takeaways
- Utilize SynthID for watermarking AI-generated content.
- Implement C2PA Content Credentials for tracking media modifications.
- Leverage Google Search features to verify content origins.
- Engage with industry partnerships to standardize content verification.
Full Summary
As generative media becomes more prevalent, understanding the origins and modifications of content is crucial. Google is expanding its content verification tools, including SynthID, a digital watermarking technology that embeds signals into AI-generated media, and C2PA Content Credentials, which track how media is created and altered.
Over 100 billion images and videos have been watermarked using SynthID, and this technology is now integrated into devices like the Pixel 10 smartphone. Users can verify content through the Gemini app and Google Search, with features that identify AI-generated media.
Partnerships with companies like OpenAI and Meta aim to standardize these verification tools across platforms, enhancing trust in digital content.