In the summer of 1955, John McCarthy and three collaborators proposed a workshop at Dartmouth for the following year. They called it the Dartmouth Summer Research Project on Artificial Intelligence. McCarthy later said that one reason for inventing the term was to escape association with cybernetics, the field Norbert Wiener had named in 1948 for the study of control and communication in the animal and the machine. Cybernetics was about feedback and steering. It put nerves, thermostats, organizations, and machines inside one field of inquiry. Artificial intelligence drew a boundary around a new one.
We have been living inside that boundary for seventy years. “Artificial” makes the system sound separate from the human world that built it. A thing outside the human world seems to owe nobody anything. It cannot be a neighbor, a tenant, a licensee, or a defendant. It appears as an autonomous object to be owned, feared, or banned.
I want to propose a different civic word: assistive. This is not a claim that every technology classified as artificial intelligence is benevolent, accessible, or designed as assistive technology. It is a claim about relationship and accountability. Contemporary systems are designed by people, trained or programmed through human choices, operated on physical infrastructure, and deployed toward goals selected by institutions. Calling them assistive returns the human being to the center of the system and forces a public question that artificial avoids: who is this helping, on whose terms, and what power remains with the person receiving the help?
That is a harder claim about the technology than vilification is. A monster has no owner. A deployed assistive system has a person it claims to help and a chain of developers, vendors, purchasers, regulators, insurers, landlords, utilities, and public officials responsible for how that help is defined. They can be named, audited, taxed, zoned, subpoenaed, and outvoted. The person can also be heard, believed, equipped, and given the power to refuse.
The rights are moving up
On December 11, 2025, President Donald Trump signed Executive Order 14365, “Ensuring a National Policy Framework for Artificial Intelligence.” It directed the attorney general to establish an AI Litigation Task Force with responsibility for challenging state AI laws that the administration considers unconstitutional, preempted, or incompatible with federal policy. It also directed the Commerce Department to identify “onerous” state laws, including laws that require models to alter what the order calls truthful outputs or that compel disclosures alleged to violate the First Amendment. A separate provision contemplated restrictions on states’ access to nondeployment funds under the Broadband Equity, Access, and Deployment program.
Read that with the word assistive in mind. A labeling rule can become a compelled-speech dispute. A bias-mitigation rule can become a preemption dispute. The First Amendment, which most Americans picture as a shield held by a person, can be carried into court by a company seeking to limit a state’s power to govern a technology deployed within its borders. The company arrives as a rights bearer. The person the system claims to assist may still arrive without a usable explanation, an inspectable record, or an appeal that can stop the harm.
The private version arrived in Minnesota. On July 27, 2026, xAI sued to block HF 1606, a state law aimed at technology that generates fake nude images of identifiable people. xAI argued that the law burdened protected expression and swept too broadly. Minnesota answered that the statute targeted a specific technology associated with nonconsensual sexual imagery. Courts denied xAI’s request for a temporary restraining order and, on September 4, its request for a preliminary injunction. The litigation continues.
That procedural history matters. Corporate speech claims do not automatically win. Yet the structure of the dispute remains instructive: the company had the resources to turn a state rule into a federal constitutional case before enforcement began. Most households experience automated judgment from the opposite direction, after a benefit, opportunity, or measure of liberty has already been taken from them.
States continued legislating. Tech Policy Press counted 109 state AI laws and 28 state data-center laws enacted by July 1, 2026. That count is evidence of democratic activity. The executive order is evidence that much of that activity was placed under organized federal scrutiny.
The rights are not moving down at the same speed
In 2024, Chief U.S. District Judge Waverly Crenshaw ruled that Tennessee’s Medicaid program had violated federal Medicaid law, the Due Process Clause, and the Americans with Disabilities Act. The case concerned TennCare policies and systems that terminated or obstructed coverage for eligible people, including notices that did not adequately explain the basis for termination and processes that failed to provide meaningful hearings. Crenshaw wrote that poor, disabled, and otherwise disadvantaged Tennesseans should not need luck, perseverance, or zealous lawyering to receive benefits to which they are legally entitled.
Other cases reveal the same accountability problem through different technologies. Arkansas replaced nurses’ assessments with an algorithm that reduced attendant-care hours by an average of 43 percent among affected plaintiffs. Idaho reduced individual Medicaid budgets through a methodology the state initially resisted disclosing; federal litigation produced due-process protections and greater access to the basis of those decisions. In Wisconsin, the state supreme court permitted limited use of the proprietary COMPAS risk assessment at sentencing, even though the defendant could not inspect the full methodology, while requiring courts to observe warnings about its limits.
These cases do not establish that every automated decision is unlawful or that trade-secret claims always prevail. They establish something more useful: when public power is exercised through a technical system, ordinary rights to notice, explanation, and a meaningful opportunity to challenge the decision become harder to use. A person cannot test the government’s story without access to the facts and rules that produced it.
Goldberg v. Kelly recognized in 1970 that eligible recipients must receive notice and an opportunity to be heard before certain public benefits are terminated. That principle survives. What the United States still lacks is a general federal right requiring every consequential automated system to identify itself, disclose an intelligible basis for its output, provide the affected person’s relevant record, and support a practical appeal. Some constitutions, statutes, and state laws supply pieces of that protection. The household faces a patchwork; the deploying institution hires counsel.
The asymmetry is therefore visible. Speech protections may be asserted upward by companies contesting regulation. Proprietary protections may be asserted downward when people seek the basis of a decision. Between them sits a household trying to learn what happened.
Why vilification makes the problem harder
I understand the anger toward AI. Many underlying complaints are justified. Vilification, however, is a poor substitute for governance. It can work in owners’ favor in three ways.
First, panic becomes a licensing regime. If these systems are described as beyond ordinary human competence, the institutions with the most capital gain the strongest claim to custody. Safety then becomes something purchased from the same concentrated firms whose power produced the concern.
Second, monsters produce broad statutes. A law aimed at a category of evil will tend to be vague or sweeping. That breadth creates constitutional and preemption vulnerabilities. A rule stating that a landlord may not deny an application on the basis of a score the applicant cannot review is narrower, more legible, and more defensible than a command to prevent harmful artificial intelligence.
Third, vilification replaces ownership with character judgment. Residents who ask about water, electricity, noise, tax abatements, and emergency planning are labeled fearful or anti-growth. People who use the systems are labeled lazy, deluded, or sick. Neither charge answers who owns the compute, who controls the data, who bears the cost, or who can contest the output. A city can survive disagreement. It cannot govern by diagnosing everyone who disagrees.
What Indiana is actually doing
Indiana has become a field site for this argument. Citizens Action Coalition reports that approximately sixty large hyperscale data centers have been proposed in the state over the past two years. Indiana University’s Environmental Resilience Institute reported that, as of July 2026, eleven counties had adopted data-center ordinances, eighteen had imposed moratoria, and two had enacted bans.
Indianapolis moved as well. On August 10, the City-County Council voted 23 to 1 for a moratorium on new data-center development through the end of 2027. The Metropolitan Development Commission gave final approval on August 19. Three previously approved projects remained exempt: DC Blox in Warren Township, Metrobloks in Martindale-Brightwood, and Sabey in Decatur Township.
The energy consequences are already entangled with federal policy. Department of Energy emergency orders have kept aging coal units available beyond planned retirement dates, including units in Indiana and elsewhere in the Midwest. The department’s orders cite projected demand that includes data centers and AI. Utilities, regulators, lawmakers, and advocates continue to dispute the reliability rationale, legal authority, and allocation of hundreds of millions of dollars in resulting costs.
Public opposition is broad, though its measured size depends on the survey. Gallup reported in 2026 that 71 percent of Americans opposed construction of an AI data center in their local area, including 48 percent who strongly opposed it. A Reuters/Ipsos poll produced a lower majority. The durable finding is not unanimity. It is that local resistance crosses ordinary partisan boundaries.
The Indianapolis moratorium was a defensible pause. By itself, it remains a defensive act about land use. It does not create a resident’s right to inspect a consequential model, access locally governed compute, or hold a share in the infrastructure that processes neighborhood data. When the pause expires, firms with capital will arrive with plans. The intervening time should be used to build a public plan.
The other plan
In Mapleton-Fall Creek, Joy Repair is developing a neighborhood-scale assistive-capacity concept: locally governed computing, a resident-controlled neighborhood census, privacy and security rules written with the people represented in the data, and outputs labeled by confidence so that a finding cannot quietly masquerade as a fact. The measure is not whether the system appears intelligent. The measure is whether a neighbor gains communication, evidence, agency, and usable power.
The physical design remains a prototype, not a finished engineering claim. Waterless or near-waterless cooling architectures exist, but feasibility depends on computing load, equipment density, climate, redundancy, power source, capital cost, and the method used to reject heat. An earth-sheltered structure and stable ground temperatures may contribute to a design; they do not, by themselves, prove that passive cooling can handle a useful computing load. Per-resident cost and the economic value retained from neighborhood data also require transparent models before they should be published as facts.
That discipline strengthens the political idea. The vault is not the point. The constitutional posture is. A household with a legally protected share in local assistive capacity enters the argument as a participant instead of merely appearing as a subject in someone else’s dataset. It may be able to communicate in forms institutions recognize, generate evidence, test a score, document a neighborhood condition, and arrive at a hearing with more than a denial letter.
Ivan Illich argued that tools can cross a watershed: they can enlarge people’s ability to act, then become systems that determine what people are permitted to do. At that watershed, the civic question is whether the tool remains available to the people who must live with its consequences. We are standing at that point with systems whose artificiality has become an alibi for concentrated ownership.
One test
Apply one test to every AI bill, ordinance, procurement, and executive order that crosses your desk in the next two years:
Does it create a right a household can exercise,
or only a duty a company can absorb?
Notice written in plain language. Access to the relevant record. A right to a meaningful explanation. An appeal that pauses the harm when delay would make victory useless. Independent audit authority. Public or cooperative access to compute. These are powers a household can use.
Vague safety mandates, voluntary ethics frameworks, and advisory boards may still have value, but they are easily converted into compliance programs. A large company can price them, staff them, and move on. The household test asks what remains in the hands of the person when the meeting ends.
The machine is not independent of us. It assists through the people, records, incentives, and institutions that organize its use. That is the danger, and it is the opportunity. Assistance can enlarge a person’s agency, or it can become the polite name for dependency, surveillance, and control. Give it to a monopoly and it will assist the monopoly. Build it with a neighborhood, and it can answer to the people who live there.