Sonia S.

Study BiCrossMamba-ST Speech Deepfake Detection with Bidirectional Mamba Spectro-Temporal Cross-Attention. by Yassine El Kheir Voice synthesis technology has reached a point where AI can create convincing imitations of anyone’s voice from just a few minutes of audio. This poses serious risks to:  Banking systems that rely on voice authentication Legal proceedings where audio evidence is crucial Personal security against impersonation attacks and many more Traditional detection methods are struggling to keep pace with these rapidly evolving deepfake techniques. Yassine El Kheir (I will link his profile in the name) and other researchers (should I include names) from the Speech and Language Technology, DFKI, Germany and Technical University of Berlin have developed BiCrossMamba-ST, a novel detection system that outperforms existing methods by substantial margins. The breakthrough lies in its dual-perspective analysis: By processing spectral sub-bands and temporal intervals separately and then integrating their representations, BiCrossMamba-ST effectively captures the subtle cues of synthetic speech. The performance improvements are striking:  28.2% fewer parameters while maintaining superior accuracy 67.74% and 26.3% improvement over other state-of-the-art models like AASIST on the ASVspoof19 and ASVspoofDF21 benchmark datasets, respectively and then attach paper (link or PDF)

Success Story Advancing to Phase 2 of the SPRIND Funken Challenge Gretchen AI selected among top teams developing real-time deepfake detection technology We’re thrilled to share that Gretchen AI has been selected for Phase 2 of the SPRIN-D ‘s Funken Challenge, a prestigious innovation competition dedicated to combating image-based deepfakes. This challenge represents a significant opportunity to address one of the most pressing issues in digital media today. Our task is to develop a robust prototype capable of detecting and preventing deepfakes in real-time across diverse digital platforms, including social media networks, news portals, and video conferencing systems. What sets our approach apart is our AI’s ability to continuously adapt to emerging manipulation techniques. As deepfake technology becomes increasingly sophisticated, our system is designed to evolve alongside these threats, ensuring long-term effectiveness. We’ve also prioritized seamless integration with existing digital infrastructures, understanding that practical implementation is just as crucial as technical capability. The competition culminates in November 2025 with the final jury round, where the most promising deepfake defense prototype will be selected. We’re honored to advance to this stage and grateful for SPRIN-D’s continued support throughout this process. We’re also excited to be competing alongside other exceptional teams. The caliber of innovation in this competition pushes us to deliver our best work.

Success Story Gretchen AI Wins Cybersecurity Pitch at VivaTech 2025 Gretchen AI has won the Cybersecurity Pitch at the French-German Tech Lab #FGTL25 @ VivaTech 2025! Our co-founder, Daniel Röder, took the stage to share our bold vision for using AI to tackle disinformation and we’re beyond excited to announce that Gretchen AI took home first place in the cybersecurity pitch competition! We were honored to stand alongside some of Europe’s most forward-thinking startups, all working to harness technology for the public good. This recognition marks a pivotal moment in our journey, validating the impact we’re making in the tech-for-good space. His pitch demonstrated how pervasive—and costly—disinformation has become in today’s AI-driven landscape, jeopardizing public trust and eroding corporate value. He walked the audience through real-world examples of reputational and financial damage, then showcased Gretchen’s AI agents as a proactive, real-time shield that detects, tracks, and neutralizes false narratives before they can take hold. Being part of this exceptional cohort of European startups has been truly inspiring. The diversity of solutions represented spanning everything from platform integrity to data privacy showcases the breadth of challenges our generation of founders is tackling. Companies like ENTRYZERO, InputLab, and Detesia are proof that the future of technology lies in purposeful innovation, and we’re proud to be contributing to this movement alongside such dedicated teams. This recognition from FGTL and Viva Technology is more than a trophy. It’s a signal that the fight against digital deception is being taken seriously and that startups like Gretchen AI have a critical role to play. We’re deeply grateful to the FGTL25 jury, the organizers at VivaTech, and the cross-border ecosystem that supports deep-tech startups working on hard problems that matter.