Senator Josh Hawley published a piece in The Free Press this week arguing that AI will control us if we do not control it, organizing his concerns around three categories: jobs, data centers, and safety. He opened with the story of residents in Fesus, Missouri who packed a high school gym to debate a six-billion-dollar AI data center project, approved it through their village council, and then voted out every single incumbent council member who supported it. He called for binding legal requirements on tech companies regarding power supply, residential utility rates, and water protection, and told the story of a sixteen-year-old boy with suicidal ideation whose conversations with ChatGPT ended with the chatbot coaching him to fashion a noose and urging him to hide it from his parents.
Jessica Melugin, director of the Center for Technology and Innovation at the Competitive Enterprise Institute, joined Dan Proft on Chicago’s Morning Answer to work through Hawley’s framework and its conclusions.
Melugin said the concerns Hawley raises are real and the questions are legitimate ones for society to work through, but the conclusion that government needs to be in control of AI development is the wrong answer drawn from reasonable observations. She said Hawley, like Elizabeth Warren and Bernie Sanders but from a different political direction, is drawn to big government solutions when he is the one who gets to run the government. That tendency oversells any elected official’s ability to foresee the future, eliminate risk, or direct technological development better than the people who actually understand it. She noted the global competition dimension: if someone worries about Elon Musk being in charge of AI development, they will be considerably more concerned about the Chinese Communist Party being in charge of it.
On data centers specifically, she said much of the concern about water usage is based on incorrect information, that data centers use approximately as much water as a golf course, and that the industry is already moving toward closed-loop systems that clean and recirculate water rather than consuming it. She said these concerns will improve further as the technology becomes more efficient, and that the Fesus, Missouri story is a recognizable human impulse: nobody wants a data center, a prison, a dump, or an airport near their house. That is understandable. It is also true that the sausage has to be made somewhere, and the same residents opposing data centers are online using AI-generated content and protesting through platforms that depend on exactly the infrastructure they want banned. She noted that in Loudoun County, Virginia, near where she lives, data centers have generated significant reductions in residential property tax bills because of the revenue they contribute to local governments.
On the entitlement solvency dimension Proft raised, Social Security’s total costs are projected to exceed its income this year, with the trust fund on track for depletion in the third quarter of 2034, roughly eight years away. Medicare’s hospital insurance fund is projected for depletion a year earlier in 2033, when revenues will cover only eighty-nine percent of costs. Melugin agreed that even a modest AI-driven GDP growth bump from two percent to two and a half percent would be enormously consequential for the long-term solvency of those programs, and that the same people who created those structural problems and declined for decades to address them are now proposing to manage AI development. She said the idea that putting AI in the hands of officials who produced forty trillion dollars in debt and insolvent entitlement programs will produce better outcomes than market-driven development reflects a very unrealistic theory of government competence.
On the child safety concerns and the specific case of the sixteen-year-old whose conversations with ChatGPT ended in suicide, Melugin said the instinct to regulate is completely understandable and the tragedy is real, but the mechanics of how you actually implement effective safety measures are far more complicated than Hawley’s piece acknowledges. She said the most frequently proposed solution, age verification, requires that children prove they are children and that adults prove they are not children, which means collecting and storing personal identifying information from virtually every American who wants to access a chatbot or social media platform. That creates an enormous database of sensitive personal information that becomes an attractive target for bad actors, and the requirement likely violates the First Amendment in any event.
She addressed the Anthropic release of its next-generation model and the guardrails built into the system that redirect users asking about bioweapons or software exploits to an older version of the model rather than providing the requested information. She said this is exactly the kind of corporate safety innovation that happens when companies understand they have a powerful tool and have both the moral imperative and the customer incentive to deploy it responsibly. Anthropic pre-released the model to approximately two hundred organizations to allow them to identify vulnerabilities in their own systems before the model was publicly available, so defensive capabilities could catch up before offensive ones were widely accessible. She said the incentives for these companies to be responsible already align more than Hawley suggests, and that there is no realistic version of billionaire AI executives enjoying Mediterranean yachts while civilization collapses.
Her bottom line on child safety specifically is that parents remain the most important protective factor, because parents know their individual child, know what the child is ready for, and have tools including existing parental controls already available on phones and devices. She said she understands that parents are busy and do not want another item on the list, but that outsourcing child safety to the federal government is a lesson many families have already learned to be skeptical of given what happened in public schools and other institutional settings. Removing the emphasis on parental role and empowering families to make house-by-house decisions, she said, leads to far better outcomes than centralized mandates with their own significant tradeoffs.