Preliminary injunctions turn on equity rather than merits, and Judge Donovan Frank's memorandum spends its weight there: a balance of harms in which, in his phrase, the public interest tips steeply in favor of the State, weighed against a plaintiff that waited roughly three months after passage and then filed five days before the effective date [2][5][5]. He was explicit that the underlying First Amendment questions are complex in the context of a new technology, and expected them to be worked out as the case proceeds [4]. What providers are pricing today is enforceability during litigation, which arrives sooner than any doctrine would.
The element Minnesota deleted is the one the rest of this area keeps. The federal Take It Down Act and a 2025 Texas deepfake approach both hang on whether the maker knew the image was nonconsensual, and HF 1606 does not [10], so liability attaches to what the product will do when a user asks, at up to $500,000 per violation, with a damages claim by the person depicted stacked on top [8][9].
The ceiling arithmetic makes the scale of exposure plain. A report referenced in the ruling counts about 3 million sexualized images from Grok in the 11 days after an image-editing feature launched, some 273,000 a day, of which about 23,000 appeared to depict children [13][3]; the smaller number multiplied by the statutory maximum is $11.5bn, and the full 3 million reaches $1.5 trillion [2][4]. Neither is a real exposure, because the reported total is not broken down by whether the depicted person was identifiable or in Minnesota, which is the element the statute requires [8][7]. What it does establish is that per-image pricing against that output rate makes any non-zero enforcement rate unbounded, which is why xAI's own fallback in earlier argument was to suspend the features for Minnesota users rather than pay [16].
Switching a feature off in one state is cheap, yet xAI is instead spending on an appeal to the 8th Circuit [15], which means the thing being defended is national capability rather than Minnesota revenue. xAI is defending against the statute being copied, because a bill that cleared the House 132-1 and the Senate 65-0, 197 votes to one, after reports of a man fabricating sexual images of more than 80 women he knew [11][1], is not a hard act for another legislature to follow.
The exposure is real now, the precedent is not yet set, and three paths lead out from here. The 8th Circuit stays enforcement and the intent-free standard dies before anyone imitates it; it lets the law stand and image-model availability goes state by state; or a merits ruling reads "letting users generate" as a duty of design and safeguards, which puts liability back on operator conduct and undoes most of the capability framing Cryptopolitan reads into the order [18]. The first and third would break the thesis. So would a record in which xAI's numbers, 52,222 accounts suspended this year and 73,604 reports to the National Center for Missing and Exploited Children, about 1.41 reports per suspension [14][6], come to count as compliance, which is precisely what Attorney General Keith Ellison's position denies when he credits the law itself with stopping products like Grok Imagine from generating sexualized images [12].
