Microsoft promises to watermark AI-generated images and videos

Pope of Balenciaga. Fake Pentagon Explosion. Distinguishing AI-generated images from the real thing is becoming increasingly difficult and sometimes disastrous.

No solution has yet been found. But Microsoft is giving it a try with a new media provenance feature debuting at its annual Build conference.

A new media provenance feature released for Bing Image Creator and Designer, Microsoft’s Canva-like web app that allows you to generate designs such as presentations and posters and share them on social media and other channels, allows consumers to create images or You should now be able to see if the video was generated. Microsoft says it’s due to AI. Scheduled to roll out in the coming months, the feature uses cryptographic techniques to mark and sign his AI-generated content with metadata about the provenance of an image or video.

It’s not as simple as a visible watermark. To read the signatures, the site must comply with the Coalition for Content Provenance and Authenticity (C2PA) interoperable specification should be adopted. Only then will sites be able to alert consumers when content is AI-generated, altered or created by designers and image creators.

So the question is, with so many image generation tools not adopting similar media provenance standards, will Microsoft’s effort make a big difference? C2PA do Backed by Adobe. Adobe recently launched its own generative AI tools, including an integration with Google’s Bard chatbot. However, Stability AI, one of his most prominent players in the generative AI field, has shown no intention of adopting the type of specification Microsoft is proposing.

Standards aside, Microsoft’s move to adopt media provenance tracking mechanisms is in line with broader industry trends associated with the prevalence of generative AI. Google announced in May that it will use embedded metadata to inform visual media created by generative AI models. Separately, Shutterstock and generative AI startup MidJourney have adopted guidelines for embedding markers that indicate content was created by generative AI tools.