When AI decides

1 min read · May 1, 2026
New Power Labs

In the spring of 2025, two DOGE employees used ChatGPT to cancel more than $100 million in federal humanities grants that were deemed to conflict with the U.S. administration's agenda. They fed the chatbot short summaries of projects from the internet with a simple prompt: "Does this relate to DEI? Answer yes or no."

The results were predictable when judgment is outsourced to a tool that has none. A Holocaust documentary was flagged for "amplifying marginalized voices." A 40-volume scholarly series on American music history was "DEI." A British general's Revolutionary War papers were guilty of "promoting inclusivity." 

Agency staff had already reviewed the grants. The DOGE employees, with no background in the humanities, set those evaluations aside and used ChatGPT instead. Nearly 1,500 grants were terminated with no appeals allowed. Four major humanities organizations have since filed a joint lawsuit, arguing the cuts were unconstitutional and discriminatory.

AI becomes embedded in how institutions screen, sort, and decide in hiring, grantmaking, and investment, serving as a shortcut for any decision-maker who wants efficiency without accountability. Yet it cannot weigh context or challenge its own judgment. 

"A computer can never be held accountable, therefore a computer must never make a management decision." — IBM Training Manual, 1979.

Without meaningful oversight, algorithms exacerbate biases and blind spots with high speed and at scale.

Narinder
New Power Labs

Like what you’re reading? Subscribe to get weekly Equity Shots in your inbox.