dpa develops trusted information layer to supply AI agents with verified news
Yannick Franke, the AI Team Lead for Deutsche Presse-Agentur (dpa), announced the development of a new information layer designed specifically for AI agents during the WAN-IFRA Frankfurt AI Forum.
The initiative, titled dpa-iq, is a response to a fundamental shift in how information is consumed and processed within the media industry. For 77 years, the German news agency has operated on a dual model: providing a real-time wire service for global events and maintaining a comprehensive news hub for archival research. However, Franke noted that the emergence of AI intermediates is disrupting this framework, as knowledge work is increasingly performed by automated systems rather than human editors.
The dpa-iq platform is designed as an API-driven destination where AI agents can retrieve verified data to complete specific tasks. Unlike traditional news archives intended for human readers, this system serves as a “trusted information layer,” allowing agents to query specific topics and receive reliable materials—including text, images, audio, and B-roll video—to fulfill their operational goals.
In a practical application, a journalist could deploy an agent to aggregate recent developments regarding the political situation in Iran or locate specific visual assets of a politician in a particular setting. The agent would query dpa-iq to ensure the retrieved content is factually accurate and sourced from a professional journalistic entity.
While the platform initially relies on dpa’s own extensive archives, the agency is pursuing a multi-source retrieval strategy. Franke indicated that dpa is currently in negotiations with external partners to integrate specialized data streams. One primary focus is sports data, which often requires structured statistics rather than narrative articles. The system will incorporate structured data from various German government bodies, organized by geographical levels to provide granular administrative information.
From a technical perspective, dpa-iq is built as a modular foundation rather than a static database. This architecture allows the agency to swap vendors, technologies, and services as AI infrastructure evolves. The platform includes a multi-source retrieval endpoint for querying various databases and a generation endpoint to produce answers. Franke emphasized that the generation tool is intended as a utility for developers to build applications upon, rather than a consumer-facing chatbot.
To facilitate adoption, the system includes API management tools that allow for the definition of access rights and rate limits for multiple users. This allows internal and external product teams to use dpa-iq as a backbone for new media tools.
The agency is also integrating the service with existing AI and automation ecosystems. This includes connections to the AI integration platform Langdock and OpenAI, as well as workflow automation tools such as Zapier, n8n, and Make. A current demonstration of this capability involves an automated workflow that scans the dpa-iq archive at 6:00 a.m. Daily to compile a publication-ready newsletter based on specific criteria.
The platform remains in private preview as dpa continues discussions with external data partners to expand the system’s reach.
