SFI MediaFutures plays a central role in Project Reynir, an initiative aimed at developing technology that enables media companies to trace and verify the origins of images and other digital content. By using the new standards from the Coalition for Content Provenance and Authenticity (C2PA), we help ensure the authenticity and trustworthiness of online content. Working alongside partners such as the BBC, The New York Times, and Microsoft, MediaFutures provides the theoretical foundation for this project.
Our research consists of a comprehensive study on media trust, replicating and expanding on previous research by the BBC. This study includes participants from the UK, US, and Norway, and examines how readers perceive the trustworthiness of images and news outlets across different contexts.
We are introducing “provenance information” for images used in news articles. This provides readers with additional details about each image, such as where, when, and by whom it was captured, as well as information about any edits made. We are testing various levels of C2PA provenance, offering different amounts of data to users to help them assess the credibility of both the image and the news source.
Recognizing that trust in news outlets varies significantly, we are including outlets with both high and low levels of public trust in our study. This allows us to investigate whether an outlet’s existing reputation influences how provenance information affects readers’ perceived trust in the images and content.
Amazon joins C2PA
The C2PA just recently expanded. On September 12, 2024, Amazon officially joined the steering committee of the Coalition for Content Provenance and Authenticity (C2PA). With Amazon now on the C2PA steering board, the push for transparency in digital content takes a significant step forward.
Amazon will begin using C2PA’s Content Credentials with images generated by its Titan AI models, allowing consumers to verify the origin and authenticity of AI-created images. The company is also incorporating this technology into AWS Elemental MediaConvert, a video processing service used by broadcasters, to help news outlets and content providers confirm the authenticity of the media they share.
“All images created by Amazon’s Titan models include an invisible watermark, providing a way to curb misinformation. Joining C2PA and implementing Content Credentials is a key part of our goal to ensure AI is used safely and responsibly”, said Vasi Philomin, Amazon Web Services’ VP of Generative AI.
New results shown at MediaFutures Annual Meeting
We look forward to welcoming Vanessa Murdock, Sr. Manager Applied Science at Amazon Web Services, as she delivers a keynote speech at our MediaFutures Annual Meeting this November.
There, Murdock will talk about responsible AI, focusing on hate speech detection in the Amazon ecosystem. Her insights will help shape our ongoing collaboration and further our shared mission. We see this as a steppingstone into increased collaboration with AWS.
Another key highlight of our Annual Meeting will be the release of our Reynir study. This research will offer new insights into media trust and the impact of provenance information on news credibility. By expanding on the original BBC study, our findings will provide valuable data on how readers assess the authenticity of digital content across different regions.