Who wins the AI race is the wrong question
J Studios via Getty Images
COMMENTARY | If the debate focuses on who wins, people will be an afterthought. The real question is: who benefits?
Artificial intelligence companies have embraced a new word for slowing development when safety measures fall behind: “pacing.”
Leading frontier AI companies are pausing model work over cybersecurity concerns, releasing hundred-page reports detailing their safeguards, and calling on government to help develop tools to pace AI when its capabilities begin advancing faster than society can safely manage.
Knowing when to slow development is important. But it cannot be the full measure of responsible AI.
For most Americans, AI is not an abstract race toward the next technological frontier. It is a new tool appearing in their workplace, a data center drawing power from their local grid, an automated system influencing whether they receive a job, a loan, or a public benefit, or a chatbot giving them information that may shape an important decision.
Pacing may help society prepare for future AI capabilities. But it doesn’t answer who benefits from AI today, who bears its costs, and who carries the risk when it fails.
Instead of asking who will win the AI race, the question we should be asking is: how is AI impacting people today?
Those affected by AI deserve more than a promise that government will intervene if the technology becomes too powerful in the future. They deserve a voice in how AI systems are being designed, developed, and deployed today.
A more people-centered AI approach starts with examining who benefits from AI, who bears its costs, and who carries the risk when an AI system fails. Fortunately, several states are beginning to show what this approach looks like in practice — from studying AI’s impacts to setting up independent advisory boards to learning by doing.
Take California for example. Under Gov. Gavin Newsom, California state agencies are studying AI’s effects on employment, identifying early warning signs of workforce disruption, and examining whether certain demographic groups could be disproportionately affected. In New York, Gov. Kathy Hochul established an independent advisory board that brings together leaders from technology, labor, nonprofits, finance, health care, and civil society to advise the state on responsible AI adoption, workforce impacts, and public-interest applications.
New Jersey is learning through implementation. Its NJ AI Assistant, an internal chatbot available to state employees, has been used by roughly 20,000 workers across more than 300,000 sessions and one million prompts. And North Carolina created an AI Task Force to evaluate how state agencies use AI and help shape a statewide approach to policy and regulation that puts people first.
While these efforts differ in scope, they begin with the same principle: AI should start and end with people.
This means involving public employees before introducing tools that will change their work. It means testing systems with the people who rely on government services, especially those who have historically been excluded from the design process. It means measuring whether a tool reduces wait times, errors, abandoned applications, and unnecessary paperwork.
Yes, we need policymakers focused on the frontier of AI development. But we also need them focused on the family wondering why the electricity bill skyrocketed, the public employee learning how to navigate an unfamiliar tool, and the person interacting with an automated system to access food assistance.
As long as the debate stays about which companies win, people will remain an afterthought.
The states doing this well have simply asked a different question — not who wins the AI race, but who benefits?
That's the question worth answering.
Amanda Renteria is CEO of Code for America.