Jane Horvath Writes About Institutional AI Bias for Dow Jones Risk Journal
In the Media | July 16, 2026
In her article “Institutional AI Bias: A New Governance Risk” for the Dow Jones Risk Journal, partner Jane Horvath notes that AI models trained on proprietary institutional data develop institutional biases: “The problem isn’t that artificial intelligence gets things wrong. It is that AI gets things right in ways that are institutionally specific and undisclosed. As AI deployment scales across both public and enterprise models, so does the divergence between what different institutions’ AI systems treat as true, reasonable, and correct.”
To deal with institutional AI bias, Jane suggests three steps that are immediately actionable for any organization deploying AI: testing AI against other models, documenting where AI analysis comes from, and understanding the AI systems influencing the people and organizations that your organization interacts with.