Key Takeaways
- Northwestern University’s Generative AI + Journalism Initiative launched the Agentic AI Investigative Journalism Challenge, a global competition running from May 15 to July 15.
- The contest asks journalists, data scientists, developers, and technologists to create reusable AI “agent skills” that speed up investigative work without replacing human reporters.
- Submissions must include interaction traces (full logs of model sessions) and a README.md that documents the workflow, data sources, and any potential legal concerns.
- Judging criteria emphasize repeatability: skills should work on new datasets and different types of investigations.
- The challenge builds on a $1 million Knight Foundation grant awarded in April 2024 to develop responsible generative‑AI practices in news production.
Overview of the Challenge
Investigative reporters routinely face mountains of documents—sometimes thousands or even millions of pages—that hide the truth they seek. “That could take months, or even years, of work to accomplish,” the initiative notes. Artificial intelligence promises to accelerate this process, yet many existing tools still fall short. To bridge that gap, Northwestern’s Generative AI + Journalism Initiative is hosting the Agentic AI Investigative Journalism Challenge, inviting multidisciplinary teams to design AI agents that act as force multipliers for investigative work.
Goals and Vision
The core aim is not to supplant journalists but to augment their capabilities. Nick Diakopoulos, professor of communication studies and computer science and head of the Computational Journalism Lab, explains: “We don’t want to replace investigative journalists… The idea is to unlock the potential of these agents to support investigative journalists — to suggest leads, patterns and connections that are apparent in the documents.” By framing AI as a collaborator, the contest seeks to surface hidden insights while preserving the journalist’s judgment and ethical responsibility.
Technical Requirements and Submission Process
Participants will work with a supplied dataset comprising U.S. House and Senate lobbying disclosures and congressional press releases from 2022 through March 2026. Using models such as Claude AI, teams must develop “agent skills”—bundles of instructions and code—that can automatically sift through the material, flag relevant passages, and propose story angles. Each entry must be submitted as a replicable workflow, allowing judges to reproduce the process on the same data.
Interaction Traces and Documentation
A critical component of every submission is the interaction trace: a complete log of the model session, including inputs, tool calls, outputs, and the precise moments when human judgment intervened. This transparency enables evaluators to see how the AI behaved and where human oversight shaped the outcome. Alongside the trace, teams must provide a README.md file that serves as a brief map of the submission. The README should list the specific skills created, indicate which findings each skill supports, locate the relevant traces, note any external data used, disclose conflicts of interest, and flag any findings that suggest possible legal violations for the evaluation panel’s review.
Broader Context and Funding
The challenge emerges from a larger effort to establish responsible practices for generative AI in news production. In April 2024, the John S. and James L. Knight Foundation awarded a $1 million grant through its Press Forward program to support this initiative. The project is a collaboration between Northwestern’s School of Communication and the Medill School of Journalism, Media & Integrated Marketing Communications. This backing underscores the foundation’s commitment to fostering innovation that strengthens journalism while addressing ethical concerns.
Potential Impact and Future Directions
If successful, the contest could produce a library of open‑source agent workflows that newsrooms worldwide can adapt to their own investigations. Nick Hagar, postdoc in the Generative AI + Journalism Initiative and creator of the contest, articulates the ambition: “We want to spark a movement around building these kinds of agent workflows… Reporters need a new toolkit to speed up critical investigative reporting processes. With this contest, we hope to demonstrate the viability of AI agent workflows and foster sharing among like‑minded journalists.” Moreover, the emphasis on repeatability ensures that skills developed for the lobbying dataset could later be applied to corporate filings, environmental reports, or international treaties, expanding the toolkit’s utility across beats.
Conclusion
By marrying journalistic rigor with cutting‑edge AI agent technology, the Agentic AI Investigative Journalism Challenge aims to redefine how reporters uncover stories buried in massive document troves. The competition’s structure—clear submission guidelines, transparent interaction logs, and a focus on reusable, ethical workflows—seeks to build trust in AI‑assisted journalism while preserving the indispensable role of human judgment. As the May 15 launch approaches, the initiative invites journalists, technologists, and data scientists to join a collaborative effort that could make investigative reporting faster, more transparent, and ultimately more impactful.
Quoted material sourced from the original announcement:
- “We don’t want to replace investigative journalists,” said Nick Diakopoulos…
- “The idea is to unlock the potential of these agents to support investigative journalists — to suggest leads, patterns and connections that are apparent in the documents.”
- “We want to spark a movement around building these kinds of agent workflows,” said Nick Hagar…
- “Reporters need a new toolkit to speed up critical investigative reporting processes…”
- “Even though we are giving folks a specific data set to work with, part of the judging criteria is how repeatable the skill is in terms of being applicable to new data sets or other kinds of investigations journalists might want to pursue,” Diakopoulos said.
https://news.northwestern.edu/stories/2026/05/artificial-intelligence-investigative-journalism