[](data:image/svg+xml;charset=utf-8,%3Csvg xmlns='http%3A//www.w3.org/2000/svg' xmlns%3Axlink='http%3A//www.w3.org/1999/xlink' viewBox='0 0 10 5'%3E%3Cfilter id='b' color-interpolation-filters='sRGB'%3E%3CfeGaussianBlur stdDeviation='.5'%3E%3C/feGaussianBlur%3E%3CfeComponentTransfer%3E%3CfeFuncA type='discrete' tableValues='1 1'%3E%3C/feFuncA%3E%3C/feComponentTransfer%3E%3C/filter%3E%3Cimage filter='url(%23b)' x='0' y='0' height='100%25' width='100%25' xlink%3Ahref='data%3Aimage/jpeg;base64,/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDABALDA4MChAODQ4SERATGCgaGBYWGDEjJR0oOjM9PDkzODdASFxOQERXRTc4UG1RV19iZ2hnPk1xeXBkeFxlZ2P/2wBDARESEhgVGC8aGi9jQjhCY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2P/wAARCAAFAAoDASIAAhEBAxEB/8QAFgABAQEAAAAAAAAAAAAAAAAAAAIG/8QAHhAAAQQBBQAAAAAAAAAAAAAAAQACAxIxERQhQWH/xAAUAQEAAAAAAAAAAAAAAAAAAAAC/8QAFBEBAAAAAAAAAAAAAAAAAAAAAP/aAAwDAQACEQMRAD8AzcU0kYbWmjjgtxXlVuXdADwIiJP/2Q=='%3E%3C/image%3E%3C/svg%3E)
The spike of AI and ML integration into the digital world is leading to numerous concerns about the credibility of content. Users cannot blindly trust content creators and publishers. It becomes important to understand the source of the material and whether it has been digitally manipulated in some way. The scope of this question about digital origins is increasing exponentially. The number of deepfakes doubled over the previous year and reached 85,047 videos in December 2020.
Given this trend, it’s becoming increasingly important to to build transparency in the digital medium. This involves a means to mark content with information on its creator, origin, and recent changes – this can be done with an old friend to digital files – like metadata.
Metadata is a reference to data about other data – with synthetic media, it’s shorthand information about a digital asset. Today, metadata gives essentially a snapshot about a particular file. To understand how it works, think about visiting the library. When you want to know about a book, you look at it’s descriptive information – the author, title, date of publication, publisher, etc.
But there’s another tool to learn more about that book’s journey, which is recorded in a library card. It also contains the borrowing history. The library card adds information beyond its intrinsic characteristics. You get an additional layer of information that provides additional insights on its journey before it reaches your hands.
And there’s work underway to transform metadata so it will work the same way. You will be able to add attribution to a picture or video so that other content consumers can check its origin and track previous changes.
This gives transparency that can help any consumer of media content understand all of the same steps in a media file’s journey before they encounter it. This is especially important for synthetic media, where it’s not always clear if a file is digitally manipulated – at all – and how. Without knowing that background, the content could be used in manipulative and even threatening ways.
[](data:image/svg+xml;charset=utf-8,%3Csvg xmlns='http%3A//www.w3.org/2000/svg' xmlns%3Axlink='http%3A//www.w3.org/1999/xlink' viewBox='0 0 10 5'%3E%3Cfilter id='b' color-interpolation-filters='sRGB'%3E%3CfeGaussianBlur stdDeviation='.5'%3E%3C/feGaussianBlur%3E%3CfeComponentTransfer%3E%3CfeFuncA type='discrete' tableValues='1 1'%3E%3C/feFuncA%3E%3C/feComponentTransfer%3E%3C/filter%3E%3Cimage filter='url(%23b)' x='0' y='0' height='100%25' width='100%25' xlink%3Ahref='data%3Aimage/jpeg;base64,/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDABALDA4MChAODQ4SERATGCgaGBYWGDEjJR0oOjM9PDkzODdASFxOQERXRTc4UG1RV19iZ2hnPk1xeXBkeFxlZ2P/2wBDARESEhgVGC8aGi9jQjhCY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2P/wAARCAAFAAoDASIAAhEBAxEB/8QAFgABAQEAAAAAAAAAAAAAAAAAAAQG/8QAHRAAAQQCAwAAAAAAAAAAAAAAAQACAwQREhMxYf/EABUBAQEAAAAAAAAAAAAAAAAAAAAB/8QAFBEBAAAAAAAAAAAAAAAAAAAAAP/aAAwDAQACEQMRAD8AzUFlrKFiLhaTIANielJnwIio/9k='%3E%3C/image%3E%3C/svg%3E)
Metadata development to add these layers of information about a piece of content’s history for synthetic media is a process that is just beginning. A Deepfake Task Force Act proposed earlier this year offers a coordinated plan that explores how a digital content provenance standard can reduce the threat of deepfakes. However, it’s not the only document that coordinates synthetically originated content in the digital space.
In 2019, Adobe introduced the Content Authenticity Initiative (CAI) in collaboration with the BBC and Washington Post.
CAI members recognize that it is crucial to expand cooperation between partners and introduce a universal standard for other companies and social media. The digital world needs regulations for platforms to identify content uploaded and shared with metadata. That, in turn, will contribute to companies actively enrolling and coordinating content in the global digital ecosystem.
Getty, Microsoft, The New York Times, Twitter, as well as synthetic media companies like Synthesia and Reface, have already joined CAI. These companies are working to map out solutions regarding content attribution and changes performed to the content.
In order to develop the theoretical ground to putting metadata into practice, there are several major challenges regarding a creator’s privacy, copyright, and data storage.
Lack of information about the source of origin is one of the biggest challenges today. Still, many independent creators are actually unable to associate their name to generated content in a way that doesn’t get lost or don’t do it at all. Hiding behind anonymity is a straight road to an unsecured digital space, specifically when trust in digital content is constantly falling. And it happens all the time in the phase between when content travels from the author to consumers.