# Deepfakes for good or for bad Source: https://www.youtube.com/watch?v=sZnSWRDSBo4 Recap page: https://rapidrecap.app/video/sZnSWRDSBo4 Generated: 2025-12-11T11:03:40.405+00:00 --- ## Quick Overview The panel concluded that deepfakes pose significant threats to media integrity and democratic processes due to the rapid advancement of AI tools, emphasizing the urgent need for provenance technologies, such as TraceID by Vermillio, and robust regulatory frameworks to distinguish real from synthetic content. **Key Points:** - The panel discussed the dual nature of AI in media, recognizing its potential for creative good while acknowledging its capacity to spread misinformation and harm individuals through deepfakes. - Sam Gregory highlighted that deepfake proliferation is happening at a staggering rate, blurring the lines between reality and fabrication, especially in sensitive areas like elections and human rights documentation. - Michaela A. Ternasky-Holland showcased her artistic work utilizing generative AI to create immersive experiences, demonstrating the positive, creative potential of the technology. - Dan Neely introduced TraceID by Vermillio, a proposed solution that tracks content provenance to help verify authenticity and protect intellectual property across digital platforms. - A key challenge identified is the difficulty in proving when content is real versus when it is AI-generated or manipulated, particularly when the content isn't English-language or politically sensitive. - The panel stressed that without widespread adoption of provenance tools and regulatory action, the erosion of trust in digital media will continue, potentially leading to societal harm. - The discussion underscored the necessity of creating a robust AI ecosystem that prioritizes human values, such as truth and creative control, over unchecked technological capability. ![Screenshot at 00:03: The opening title slide announces the 'AI for Good Global Summit' and its theme: 'Identifying innovative AI applications to solve global challenges' scheduled for 8-11 July 2025 in Geneva, Switzerland.](https://ss.rapidrecap.app/screens/sZnSWRDSBo4/00-00-03.png) **Context:** The discussion took place during the 'AI for Good Global Summit' (8-11 July 2025, Geneva, Switzerland), featuring a panel moderated by Greg Williams (WIRED) titled 'Deepfakes for good or bad?'. The panel included Michaela A. Ternasky-Holland, an Emmy/Peabody-nominated director specializing in immersive technology, and Dan Neely, CEO and Co-Founder of Vermillio, a company focusing on AI authentication tools. Sam Gregory, Executive Director of WITNESS, provided context from a human rights and digital forensics perspective, framing the rapid evolution of generative AI against a backdrop of existing disinformation threats. ## Detailed Analysis The panel engaged in a deep discussion about the double-edged sword of deepfake technology, balancing its creative potential against its capacity for malicious use. Michaela A. Ternasky-Holland provided context on the creative use of AI, showing examples from her art installations like 'Face to Face' and 'Morning Light,' which explore immersive storytelling using generative AI and AR/VR technologies, emphasizing the need to connect with ancestors and explore identity. Sam Gregory highlighted the growing threat landscape, presenting real-world examples of deepfakes used for political disinformation (like the Bawumia recording) and the difficulty in detecting them, noting that the ability to discern reality from fabrication is becoming increasingly challenging. He stressed that provenance solutions need to be developed and deployed globally to maintain trust in digital media. Dan Neely introduced Vermillio's TraceID, a proposed solution designed to create guardrails for the generative AI internet by tracking content provenance. This tool aims to help creators protect their IP, check content authenticity across platforms like TikTok, YouTube, Instagram, and Facebook, and offer services ranging from free basic verification to premium threat monitoring. The consensus among the speakers was that while AI offers immense creative potential, proactive, enforceable solutions—both technological (like TraceID) and regulatory—are urgently needed to manage its inherent risks, particularly concerning political disinformation and the misuse of personal likenesses. ### Event Context - AI for Good Global Summit - 8-11 July 2025, Geneva, Switzerland - Panel: Deepfakes for good or bad? - Participants: Greg Williams (Moderator, WIRED), Michaela A. Ternasky-Holland (Artist/Director), Sam Gregory (WITNESS), Dan Neely (Vermillio, Remote) ### Creative Applications of AI - Michaela showcased AI work like 'Face to Face' using VR headsets and 'Morning Light' using generative AI to explore identity and ancestral connection. ### Deepfake Threats - Sam Gregory presented examples of malicious deepfakes, including a 'Leaked Secret Recording of Bawumia' (falsified with AI) and an Indian politician claiming audio clips were deepfakes (real but claimed AI), illustrating the confusion between real and fake content. ### TraceID Solution - Dan Neely introduced TraceID by Vermillio, a tool for content provenance tracking across major platforms (TikTok, YouTube, Instagram, Facebook) to verify authenticity and protect IP. ### Current Challenges - Gregory noted that detection tools are not yet widespread or effective enough to keep pace with content generation, leading to a disparity where it is easier to create fakes than to reliably detect them. ### Future Needs - The panel agreed on the need for robust provenance systems and clear regulatory frameworks to build trust in digital media and prevent the misuse of synthetic content, especially in areas like elections and human rights documentation. ![Screenshot at 00:03: The opening title slide for the 'AI for Good Global Summit' scheduled for 8-11 July 2025 in Geneva, Switzerland.](https://ss.rapidrecap.app/screens/sZnSWRDSBo4/00-00-03.png) ![Screenshot at 00:08: Panel introductions displayed on the screen listing Michaela A. Ternasky-Holland, Greg Williams \(Moderator\), Sam Gregory, and Dan Neely \(Remote\).](https://ss.rapidrecap.app/screens/sZnSWRDSBo4/00-00-08.png) ![Screenshot at 00:49: A slide illustrating various deepfake scenarios, labeling one example as 'Real footage' and others as 'AI generated images' and 'AI generated video from AI-gen image' to contrast reality with synthetic media.](https://ss.rapidrecap.app/screens/sZnSWRDSBo4/00-00-49.png) ![Screenshot at 04:45: A slide titled 'About the Project' describing 'The Gr\(ai\)t Debate' as a live performance experience using AI-generated political candidates in real-time debate.](https://ss.rapidrecap.app/screens/sZnSWRDSBo4/00-04-45.png) ![Screenshot at 11:55: A graphic demonstrating TraceID's function, showing a verified master track matching a 'GENERATED' track with high accuracy \(90th Percentile\) across platforms like TikTok, YouTube, Instagram, and Facebook.](https://ss.rapidrecap.app/screens/sZnSWRDSBo4/00-11-55.png)