It’s make or break time for AI labeling systems

Source: The Verge

If robust AI labeling had been in place when the swanky images of Pope Francis went viral, it might have been easier for people to recognize them as fake.
Image: via Reddit

SynthID and C2PA Content Credentials

Two emerging technologies aim to invisibly tag image, video, and audio files with provenance information:

• SynthID – an invisible watermark applied to content generated by Google AI models.
• C2PA Content Credentials – a broader standard for embedding metadata that describes the creation and modification history of digital media.

Both are being expanded to help users verify whether a piece of media was produced by AI and to combat the spread of unlabeled deepfakes.

Google I/O Announcement

During its recent I/O conference, Google announced that developers will soon be able to verify the presence of SynthID markers in images. This capability is intended to make it easier to spot AI‑generated content and to provide transparency about the origins of media created with Google’s generative models.

Further Reading

Read the full story at The Verge.