Subscribe
Sign up to stay up to date with umanitek's latest developments.
Sign up to stay up to date with umanitek's latest developments.
Reykjavik, Iceland (October 9, 2025)
In a first-of-a-kind event hosted in Reykjavik, Umanitek AG, the Swiss-based AI company developing the largest “digital immune system” to combat harmful content and mitigate the risks of artificial intelligence, together with Videntifier, a global leader in forensic content identification, unveiled the first live demonstration of the Umanitek Guardian Agent. The system is powered by the OriginTrail decentralized knowledge network and Videntifier’s visual recognition technology.
The gathering brought together a select group of global and Icelandic figures, including investors, institutional partners, and representatives from the technology, media, and policy sectors. The event served as an exclusive preview of Umanitek’s progress toward creating a decentralized, privacy-preserving AI system that identifies and prevents the spread of harmful or illegal content online.
The live demo showcased a working prototype of Umanitek’s Guardian Agent – an AI Agent that identifies known illegal or deepfaked content and helps users initiate takedown requests. Built on the OriginTrail Decentralized Knowledge Graph (DKG), the system verifies content across distributed data sources without ever exposing or storing sensitive material.
By combining this with Videntifier’s forensic fingerprinting – trusted by law enforcement worldwide – the Guardian Agent can detect and assess content with high speed and accuracy, working solely on encrypted photo fingerprints. This ensures a scalable, privacy-first solution for those facing the growing risks of AI-generated misuse. Platforms and institutions can easily adopt the same technology to enhance digital safety while maintaining complete data control.
The Reykjavik event marks a key milestone in Umanitek’s roadmap. The Guardian Agent is now live for early adopters and partners who will validate the system before broader release. It also signals Umanitek’s next phase: expanding collaborations with global and regional partners committed to safer digital spaces.
The live demonstration walked the audience through a real-world scenario of detecting and flagging deepfaked content. The team used a controlled example featuring a young adult volunteer to illustrate how misuse of AI-generated content can impact vulnerable groups. . In the demo, her likeness was inserted into scenes she had never participated in, including a political protest and a commercial – all created using widely available Sora 2.
Using only four reference photos, the Guardian Agent created a secure fingerprint of the volunteer’s likeness, scanned publicly available media, and surfaced all instances where that likeness appeared online – from known, consented content to deepfakes and misused material never seen or approved before – enabling users to determine what’s legitimate and what’s not.
Watch the full demo video below – walking through this exact scenario and showcasing each feature of the Guardian Agent in action.
Branimir Rakic, Umanitek and OriginTrail co-founder and CTO, guided the audience through each step, showing how each verification creates a protected digital Knowledge Asset – that anchors identity in an immutable system that grants individuals verifiable proof and control over where and how their likeness appears online.
The demo also featured a short message from Bob Metcalfe – Turing Award winner and inventor of the Ethernet – who emphasized that today’s convergence of productivity tools, decentralized knowledge graphs, and blockchain technology, all part of the Umanitek Agents infrastructure, offers a powerful path toward addressing one of the greatest challenges of our time: protecting the truth.
The demo was followed by a panel discussion moderated by Laurie Segall – renowned journalist and media executive, founder of Mostly Human and former senior technology correspondent at CNN. Laurie is known for her in-depth interviews with tech leaders such as Tim Cook, Sam Altman, and Mark Zuckerberg, as well as for her award-winning reporting on the impact of harmful content online, including deepfakes. She has received both Gracie and Webby Awards for her work.
She was joined by Tomaz Levak, co-founder of Umanitek and OriginTrail, and Ari Jónsson, CEO of Videntifier. The panel explored the urgent need for a digital “trust layer” in AI – the infrastructure that protects truth, verifies digital identity, and safeguards reputations in the face of accelerating synthetic media capabilities.
Tomaz emphasized that such a layer is not only essential for safety but also unlocks new opportunities, allowing AI systems to innovate responsibly with built-in safeguards. Laurie Segall highlighted the profoundly personal nature of the issue, noting that everyone is now vulnerable to content manipulation as powerful tools for creating and spreading false material become increasingly accessible.
The Guardian Agent will gradually open to more users over the coming months.To join the waitlist and receive early access to product updates, visit:
www.umanitek.ai – and click the “Subscribe” button in the top right corner.