The Right to Compute: Elevating the Standard for AI Laws

Montana’s Right to Compute requires its legislature to justify restrictions on AI and related computational tools with evidence—not judicial imagination or regulatory vibes.

Your boss declares a new policy: all correspondence with sales leads must be done by handwritten letter. He insists it’s the only way your company will stand out as more and more competitors switch to automated processes. You’re a little less confident.

You’re not even sure you know where the mail department is (or if one exists!). You suspect your intended recipients will similarly be looking for days to find what mail you manage to send their way. What’s more, your writing hasn’t been legible since high school whereas you’re quick to develop a rapport with anyone on the opposite side of a phone call. To make it all so much worse, you’re up for promotion this year but there’s a company-wide rule that no one moves up unless their sales do too.

What do you do?

You’re extremely frustrated. You joined this firm because it promised you protection from outdated, rigid policies akin to those you’ve heard about at other companies. Unlike those rival firms, your company’s mission statement touts the freedom of creativity, persistence, and passion.

Given that your boss has announced this as one of his “ironclad” policies, your only recourse is to appeal to the (hopefully) better judgement of his boss, the Vice President of Sales. You write her a letter detailing all the reasons why she ought to reverse the snail mail strategy. You even include several printed out articles demonstrating the time that’s wasted handwriting materials, the likelihood of letters being lost, and the novel, proven sales techniques being employed by your rival salespeople at the firm down the way.

On the other hand, your boss has no evidence to offer to back this ill-advised move. Surely the VP has to strike down this unsupported policy!

A few days later you receive a letter from the VP. She’s analyzed the evidence sent over and finds your arguments compelling. That said, she’s a highly deferential leader. There’s a little plaque on the wall behind her desk that says, “Those who manage least, manage best.” She acknowledges that your boss has nothing but hunches to justify this shift. But she comes up with an entirely different rationale for the new approach. The VP cites the following conceivable reasons for his policy: cybersecurity concerns will go down, salespeople may become better writers, and stress may diminish due to fewer emails.

You remind her of the stakes: your promotion is on the line (as well as the company’s overall performance, of course). She brushes those contentions aside, leaves the policy in place, and points to the plaque’s praise for light-touch management before asking you to go sharpen your pencils for all the writing that’s in store.

Sounds like a flawed system, right? The folks in positions of authority are supposed to make sure everyone adheres to similar values and works diligently and on the basis of clear evidence to carry out the firm’s objectives, right? Surely this cannot be how any functioning organization reviews binding policy, right?

This is precisely how judges review state laws that do not blatantly discriminate against a protected class or directly violate a foundational right.

This is not a joke. Under rational basis review, a state law pertaining to social welfare or economic matters will be upheld by a court so long as the judge can come up with any conceivable justification for how the policy is rationally related to some government objective.

You read that correctly—the legislature need not do its homework, the judge will make up an explanation for a given law on their behalf. In fact, the legislature is likely disincentivized from doing so because that would prompt the court to look more closely at its work. The upshot is that we’ve set up a system under which laws are presumptively constitutional, even when they may permit a state to hinder expression, limit entrepreneurial activity, and otherwise limit the scope of personal liberty.

That’s precisely what’s likely to occur in 49 states. If and when state laws dictating how AI models are developed, how they behave, and who may use them and for what ends get challenged in court, the state will likely win. Courts may push to the side the federalism principles at issue, the free expression issues at stake, and the economic liberties being confined and simply opt to treat these AI laws like any other social welfare or economic law—in short, they’re likely to presume the legislature did its homework without even asking or checking.

Not so in Montana. In the Last Best Place (a moniker celebrated by its proud residents and one I adhere to as a former Helena dweller) there’s first-rate AI policy. The state legislature there passed the Right to Compute Act, which, in relevant part, holds:

The legislature finds that the rights to acquire, possess, and protect property under Article II, section 3, of the Montana constitution, and the freedom of expression under Article II, section 7, of the Montana constitution, also embody the notion of a fundamental right to own and make use of technological tools, including computational resources.

Any restrictions placed by the government on the ability to privately own or make use of computational resources for lawful purposes must be limited to those demonstrably necessary and narrowly tailored to fulfill a compelling government interest.

Government actions that restrict the ability to privately own or make use of computational resources for lawful purposes, which infringes on citizens’ fundamental rights to property and free expression, must be limited to those demonstrably necessary and narrowly tailored to fulfill a compelling government interest.

In brief, this means that whenever the state legislates in a manner that impacts the ability of Montanans to make use of AI tools and related technologies—from the mundane to the state-of-the-art, policymakers must only rely on interventions that are “demonstrably necessary and narrowly tailored to fulfill a compelling government interest.” As a result, the legislature must do its homework and show that it has considered alternative ways to achieve its selected ends without causing as much interference to the aforementioned right. What’s more, not just any government goal may be pursued—in such cases, the legislature must assert “a compelling government interest.”

This law creates a presumption in favor of personal liberty, entrepreneurial energy, and self-expression by way of using the latest AI tools. That’s a big deal given the rate at which states are enacting AI laws and high potential for those laws to unnecessarily burden American use of these incredible tools.

Other states can and should follow Montana’s lead. The Right to Compute correctly tips the scales in favor of legislative humility at a time when policymaking should be driven by evidence, not vibes. This Right prevents courts from indulging in speculation about what a legislature might have meant or might have been trying to achieve, such as “protecting kids” or “reducing bias.” It requires lawmakers to identify a compelling objective and to demonstrate that any restriction on computational tools is necessary and narrowly tailored to achieve it. That recalibration matters.

AI systems, as with prior generations of computing tools, increasingly function as instruments of speech, research assistants, design platforms, and engines of small-business formation. A statute that limits who may access or deploy those systems is not a minor economic regulation. It shapes the practical boundaries of expression and enterprise. Many of these laws are also likely to have extraterritorial effects, thereby subjecting residents of one state to the laws of another in violation of foundational constitutional principles. A legal regime that demands evidence and precision before curtailing those tools better reflects the scale of what is at stake.

The broader lesson is institutional. Courts need not abandon deference across the board. But when a state ventures into domains that touch the architecture of computation and communication, reflexive rational basis review is a poor fit. Montana has recognized that reality and embedded a higher standard in its own law. Other states can do the same. If they decline, they should at least be candid about the consequence: policies that restrict AI development and use will survive so long as a judge can hypothesize a reason to sustain them. That may be sufficient for handwritten letters. It is insufficient for governing the infrastructure of the modern economy.