Seattle-headquartered nonprofit TrueMedia unveiled a complimentary AI-driven media authentication tool on Tuesday to aid journalists and fact-checkers in identifying deepfakes and countering misinformation in anticipation of the forthcoming U.S. and global elections.
The bipartisan institution, established in January, is spearheaded by Oren Etzioni, a prominent AI expert and professor at the University of Washington, with financial backing from Garrett Camp, the co-founder of Uber, channeled through his philanthropic entity, Camp.org.
Oren Etzioni is the founder of TrueMedia. (Allen Institute for AI Photo)
Despite not being flawless, the tool boasts an exceptionally high accuracy rate in detecting deepfakes, standing at approximately 90% across various media formats such as images, videos, and audio, as confirmed by Etzioni. TrueMedia leverages a blend of in-house technological advancements and AI tools from its partners to scrutinize media content and generate a likelihood score indicating its authenticity.
For instance, the tool swiftly flagged a notorious fake video allegedly depicting Ukraine’s top security official confessing to orchestrating the terrorist incident at a Russian music venue on March 22 as “highly suspicious.” It unequivocally identified the video as containing AI-manipulated visuals with 100% certainty.
“If it’s a deepfake, we are highly probable to detect it,” remarked Etzioni.
In conjunction with the tool’s launch on Tuesday, TrueMedia formalized an agreement with Microsoft to exchange data and resources, fostering collaboration on diverse AI models and methodologies.
Among TrueMedia’s collaborative partners are Hive, Clarity, Reality Defender, OctoAI, AIorNot.com, and Sensity.
The tool’s debut garnered coverage from The New York Times, which highlighted instances like a counterfeit image of Etzioni in a hospital setting, generated using an AI tool. The publication underscored that such images could potentially sway an election if deployed against figures like Mr. Biden or former President Donald J. Trump just before the election.
The primary objective is to thwart such scenarios by equipping journalists and fact-checkers with TrueMedia’s verification tool, enabling them to promptly debunk fabricated content, emphasized Etzioni in an interview with GeekWire. This initiative holds significance not only for the impending U.S. presidential election but also for electoral processes in Europe, India, and other regions.
To safeguard its methodologies and prevent adversaries from circumventing detection of their deepfakes, TrueMedia is regulating access to the tool. Nonetheless, authorized users can disseminate links to content assessments on the TrueMedia website.
Previous Story
The nascent nonpartisan AI nonprofit TrueMedia, under the leadership of Oren Etzioni, is pioneering a political deepfake identification system.
“We must ensure widespread access to the tool,” articulated Etzioni, underscoring TrueMedia’s commitment to facilitating broad usage of the tool while upholding confidentiality.
Looking ahead, Etzioni revealed TrueMedia’s plans to seamlessly integrate the tool into web browsers through a browser extension and social media platforms like Twitter and Reddit, enhancing its accessibility.
With a background as the former CEO of the Allen Institute for AI in Seattle, Etzioni has been deeply immersed in artificial intelligence throughout his career, predating the surge of generative AI into popular discourse. He noted the unprecedented pace at which AI technologies have evolved recently, describing the landscape as an ongoing arms race with advancements on both ends. Despite the rapid developments, Etzioni expressed confidence in the efficacy of detection mechanisms thus far.