Advancing the Future of Generative AI Watermarking: DeepMark at ICLR 2025

DeepMark is pleased to announce our participation in the ICLR 2025 Workshop on Generative AI Watermarking, where we had the opportunity to present our latest research in the field. Representing our leadership team at the event were Slavko Kovačević (Chief Executive Officer), Kosta Pavlović (Chief Science Officer), and Murilo Silvestre (Chief Technology Officer), who engaged with the international research community on the evolving landscape of digital watermarking for generative models.

At the workshop, our team presented a new benchmarking framework designed to rigorously evaluate generative AI watermarking algorithms. As generative models become increasingly integrated into digital content creation pipelines, the need for reliable watermarking solutions capable of ensuring provenance, authenticity, and accountability has never been more urgent. Our research proposes a comprehensive set of metrics and evaluation protocols aimed at establishing a standardized benchmark for the community. This contribution is intended to foster consistency and comparability across future innovations in the space.

Our team made it possible

We are particularly proud of the collaborative effort that made this contribution possible. Special recognition goes to our newest team member, Petar Nedić, whose technical expertise and dedication played a pivotal role in the development of this paper.

We extend our gratitude to the workshop organizers, keynote speakers, and fellow panelists for facilitating an intellectually stimulating and well-organized event. The opportunity to meet and engage in direct dialogue with leading researchers and competing teams was both valuable and motivating. We deeply respect the work being done across the field and look forward to continued collaboration and healthy competition.

Open Source Initiative

To explore the details of our benchmarking study, we invite you to read the full paper, available here

At DeepMark, we remain committed to advancing responsible innovation in generative AI. This milestone is another step toward shaping robust and transparent digital ecosystems.