Northwestern Launches AI Challenge for Investigative Journalism

Investigative journalists often obtain thousands, if not millions, of pages of documents and take on the challenge of finding the truth buried inside. That could take months, or even years, of work to accomplish. Now, a new contest led by the Generative AI + Journalism Initiative at Northwestern University is working to unlock the potential of coding agents to increase the processing power of AI.

The Agentic AI Investigative Journalism Challenge is a global competition inviting journalists, data scientists, developers and technologists to build AI "agent skills," or bundles of instructions and code, to help make investigative reporting faster, cheaper and more transparent. The contest launches Friday, May 15 and ends July 15, 2026.

According to the official Northwestern announcement, competing teams will use Claude AI to find newsworthy insights in a provided dataset comprised of U.S. House and Senate lobbying disclosures and congressional press releases from 2022 through March 2026. Teams must enter their process as a replicable workflow.

"We don't want to replace investigative journalists," said Nick Diakopoulos, professor in communication studies and computer science. Diakopoulos leads the Computational Journalism Lab in Northwestern's School of Communication. "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."

Teams will submit four distinct deliverables. Agent skills represent the reusable workflow they built, including any new skills developed. A findings report provides a written summary of the newsworthy discoveries produced by running the skill(s) on the data plus any outside material brought in to enhance the investigation. Interaction traces include full logs of the model sessions, including inputs, tool calls, outputs, and the moments when human judgment intervened. A README.md serves as a brief map of the submission, including skills, which findings they support, where the relevant traces are located, any outside data used, any conflicts of interest and whether findings suggest possible legal violations that should be flagged to the evaluation panel.

The physical reality of this work involves clicking through model interfaces, watching progress bars crawl across screens, and manually verifying outputs that look suspiciously confident. (Anyone who's tried to debug an AI hallucination knows the frustration.) The interaction traces requirement forces teams to document exactly where human intervention occurred — a crucial transparency measure in an era where AI outputs often appear seamless but hide significant editorial decisions.

"We want to spark a movement around building these kinds of agent workflows," said Nick Hagar, postdoc in Northwestern's Generative AI + Journalism Initiative and creator of the contest. "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."

The top team will win $5,000, second place $2,500 and third place $1,000. All three teams will be invited to present at the 2026 Computation + Journalism Symposium. Even though the organizers 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.

The challenge emerges from an effort to develop responsible practices for generative AI in news production, which launched in April 2024 with a $1 million grant from the John S. and James L. Knight Foundation through its Press Forward program. The project is a collaboration between Northwestern's School of Communication and the Medill School of Journalism, Media & Integrated Marketing Communications.

This approach represents a pragmatic middle ground in the AI-journalism debate. Rather than treating AI as a replacement for reporters, the contest frames it as a force multiplier for document-heavy investigations. The emphasis on interaction traces and human intervention logs acknowledges that AI tools still have shortcomings that require editorial oversight.

Whether newsrooms actually adopt these workflows beyond the contest remains the real question. Many investigative teams operate with tight budgets and limited technical expertise. The prize money is modest compared to the resources needed for sustained implementation. More importantly, the contest assumes journalists have the time to learn new agent workflows while maintaining their existing reporting responsibilities.

The tools might work. The question is whether the industry will invest in them when the margins are already razor-thin.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn