"Our solution automates complex search and comparison processes, saves time and brings fact checks to where they are needed."

Disinformation and deepfakes are one of the biggest challenges facing our society. The young start-up Gretchen AI and its founding team, consisting of CEO and senior AI expert Dr Tim Polzehl, deepfake and DevOps expert Arnab Das and AI agent / RAG expert Daniel Röder, are tackling this challenge with state-of-the-art AI technology. The Berlin-based company was recently honoured with the Deepfake Detection Challenge Award from SPRIND and develops intelligent solutions for verifying media content.

In an interview with CEO Tim Polzehl, we talk about the significance of this award, Gretchen AI's unique technologies and how its AI agents help to recognise manipulation in speech, images and videos. He also provides exciting insights into the ‘Gretchen question’, which helps media houses and companies to make fact checks more efficient and accessible.

How does Gretchen AI differ from other solutions? How do you ensure that your AI does not produce misinformation? And why is Berlin the ideal location for a company dedicated to fighting disinformation? He answers these and many other questions in an interview with #ai_berlin.

Gretchen AI were recently honoured with the Deepfake Detection Challenge Award from SPRIND. Congratulations! What does this recognition mean to you? How could it help you fight disinformation even more effectively in the future?

The SPRIND award is a true milestone for us! Deepfakes and disinformation are major issues right now—especially after the recent elections. AI-powered solutions for detecting and combating these manipulations are more important than ever, and that’s exactly what we are passionately working on.

Thanks to SPRIND, we can not only further develop our technology and turn it into a real product but also benefit from a strong network. Engaging with other top innovators provides us with valuable insights and helps us bring our solution into practice even faster.

Our goal: With our Gretchen AI Agents, we aim not only to strengthen journalism but also to support companies beyond that. Whether it’s fact-checking, verification, or internal knowledge management, our AI-powered solutions help ensure reliable information and effectively counter disinformation.

Can you explain to us how your AI helps to recognise manipulation in language, images and videos? What exactly can we expect from this technology?

Gretchen AI brings the fight against disinformation into the age of AI agents! Instead of relying solely on traditional detection models, which often lag behind the latest deepfake techniques, we also focus on context. Only by understanding where an image or video originates from and the context in which it was published can we make reliable statements about its authenticity.

How does it work? Our AI agents are connected to trusted data sources and simultaneously scan the web for additional information about a piece of media—such as its origin, the people involved, the publication date, and more. At the same time, our Digital Alteration Engine detects digital manipulations and deepfakes.

This way, we provide users not only with a tool to expose deepfakes but also with a means to understand media content in its full context.

^{The founding team of Gretchen AI Dr Tim Polzehl (Senior AI Expert), Arnab Das (Deepfake and DevOps Expert) and Daniel Röder (AI Agent / RAG Expert) © Gretchen AI}

How does your ‘Gretchen question’ help users to master information verification? Could you give us a concrete example of a company that has already used this technology successfully?

With the Gretchen Question, we go straight to the core of the problem! Instead of simply classifying something as “true” or “false,” we provide the crucial context. Our Disinformation Dashboard visualizes content, giving users full control, while our Gretchen AI Agents enable interactive data analysis.

As part of the SPRIND Challenge, we are collaborating with our fact-checking role model and reference partner, dpa. Our dashboard allows fact-checkers to verify all relevant information about images – and with just one click, they can prepare a ready-to-publish fact-checking article.

Why does this matter? Fact-checking should be a much more present part of our daily lives, but the high manual effort makes it challenging. Our solution automates complex search and comparison processes, saves time, and brings fact-checks exactly where they are needed.

What distinguishes your technology from other solutions on the market, and how does Gretchen AI ensure that the answers remain truly relevant and trustworthy?

Our technology stands out by combining deepfake analysis with AI agents that not only detect manipulations but also verify context. Instead of relying solely on traditional detection methods, we offer a comprehensive framework that enables users to perform interactive and transparent fact-checking.

How do we ensure that the information is accurate and trustworthy?
First, we are collaborating with our reference partner dpa to build a reference database of verified facts. Additionally, our AI agents access existing fact-checks as well as other web resources.

Users remain in control at all times: They can review sources themselves or select only specific ones to improve the quality of the data even further.

How do media organisations and journalists specifically benefit from using your platform? Are there specific workflows that are significantly simplified by your technology?

Our Disinformation Dashboard is specifically designed to make fact-checking faster and more efficient, for both journalists and editorial teams. While traditional verification processes are often time-consuming, our platform automates many of these steps, providing real-time, context-based analyses that journalists can easily interpret and assess.

But our technology goes even further: Gretchen AI Agents can be seamlessly integrated into various editorial workflows, from fact verification and decontextualization to content analysis and content creation.

Media organizations benefit from a smart assistant that not only helps detect disinformation but also supports the fast and well-founded creation of high-quality content and counter-narratives.

How do you deal with the challenge that AI models sometimes provide incorrect or unreliable information? What measures do you take to ensure that Gretchen AI does not produce ‘hallucinations’?

Our solution to the problem of AI "hallucinations" is a smart AI agent combined with transparent source attribution. Every language model is probabilistic, and in journalism, 98% certainty isn’t enough, verifiable facts are essential.

That’s why Gretchen AI presents all information in a clear and verifiable way, always with direct source references. Our AI agents go through multiple validation cycles to ensure that the provided data is internally verified.

This way, users retain full control and can quickly assess the reliability of the information for themselves.

What advantages do you see in Berlin as a location for a company like Gretchen AI? And how does the local research landscape contribute to your work?

Berlin is the ideal location for Gretchen AI because it brings together cutting-edge research, innovation, and the media industry. Through close collaboration with research institutions like DFKI and the city’s thriving AI startup scene, we have direct access to the latest technology and expert knowledge.

At the same time, our proximity to major media organizations facilitates the development and integration of our solutions. This dynamic environment makes Berlin the perfect hub for our mission: fighting disinformation with AI.

Thanks for the conversation!