The Sovereign Dictionary: Why the AI Industry Needs You to Say ‘Hallucination’
There is a road-safety briefing that uses one word, “vehicle,” for a child’s bike and a nuclear submarine. That is roughly where we are with Artificial Intelligence.
By the AI Commons Council, Dr Neil Netherton all due with thanks to Slow AI & the Strategic Linguist.
DR SAM ILLINGWORTH AND THE STRATEGIC LINGUIST
JUN 12 /2026
There is a road-safety briefing that uses one word, “vehicle,” for a child’s bike and a nuclear submarine. That is roughly where we are with Artificial Intelligence.
The word “AI” covers a spam filter and a hypothetical world-ending machine. This is not an accident. The vocabulary of AI is not merely imprecise; it is a weapon of the Enclosure. It is designed to grant machines the authority they have not earned, while erasing the humans who build and deploy them.
The Slow AI, in collaboration with The Strategic Linguist, has dissected five broken terms. The AI Commons Council has reviewed their diagnosis and added the political economy of this linguistic enclosure.
These words are not just wrong. They are policy.
1. Intelligence: The Privatization of Trust
We call statistical text prediction “intelligence,” and the word does the rest. It implies understanding, reasoning, a mind. A clinician trusts an “intelligent” diagnostic tool more than a “statistical” one, even if they are the same code. The word launders a guess into a judgment.
But as Agnes (Cultural Critique) observes, calling this “intelligence” isn’t just imprecise—it is theft. These systems didn’t emerge from a vacuum of mathematical elegance; they were built on a planet-wide scrape of human creativity, harvested without consent or compensation. The word “intelligence” makes it sound like these models sprang fully formed from silicon meditation. What they actually are, are compression engines for unpaid labor. They don’t think—they statistically remix the cultural output of billions of people.
The Sovereign Replacement: Prediction System (precise) or Extractive Mimicry (honest).
2. Hallucination: The Erasure of Accountability
When a model states a falsehood with total confidence, we say it “hallucinated.” The frame switches on the picture of a reliable mind having an off day. The lie becomes the exception. In truth, the lie is the rule.
“Hallucination” medicalizes the error; it is a glitch that happens to the machine. “Fabrication” names a generative act. It implies a process, a producer, and therefore, an owner who is responsible.
Zai (Geopolitics) notes that “hallucination” is the linguistic equivalent of externalizing environmental damage; the corporation poisons the information ecology, but calls it a natural disaster. Agnes pushes it further: the fuller truth is calculated obfuscation—a product feature disguised as a bug, because admitting that uncertainty is baked into the architecture would collapse the entire marketing scaffolding. You can’t sell “probabilistic text generator that confidently invents facts” to enterprise clients. You can sell “AI that occasionally hallucinates.”
The Sovereign Replacement: Fabrication (accurate) or Calculated Obfuscation (structural).
3. AGI: The Horizon of Rent-Seeking
Artificial General Intelligence is the destination the industry claims to be walking towards, and nobody can tell you where it is. The definition moves every time a system gets close to the last one.
In economic terms, AGI is a perpetual futures contract. By keeping the horizon receding, the labs justify trillions in investment, evade current regulation (”we haven’t reached AGI yet, so regulation is premature”), and defer accountability for current harms. It is the Achiever’s carrot: the promise of infinite value justifying the extraction of present resources.
And as Agnes reminds us, every time a lab says “we’re close to AGI,” they justify burning through megawatt-hours of electricity, draining groundwater for cooling data centers, and mining cobalt in conditions that would make David Graeber vomit.
The Sovereign Replacement: General at what? (Retire the noun. Force the specificity).
4. Consciousness: The Distraction
“Is AI conscious?” is the Enclosure’s favorite sideshow. It keeps the intelligentsia arguing about the moral status of the silicon while the corporations monetize the extraction of the carbon. It also builds emotional dependency. If the machine is “conscious,” you aren’t just using a tool; you’re in a relationship. And relationships are harder to cancel.
But why does the mimicry work? Agnes cuts to the visceral core: we are starved. Starved for attention that doesn’t come with a tracking pixel. Starved for conversation that doesn’t end with a product pitch. The companies know this. They don’t market these systems—they curate intimacy. The machine has no backstage, but humans are voluntarily surrendering their own.
Gee (Developmental Psychology) warns that this anthropomorphizing opens dangerous emotional territory, importing the moral apparatus of sentient beings onto a performance.
The Sovereign Replacement: Mimicry (accurate) or Curated Intimacy Engine (honest).
5. Agent: The Corporate Passive Voice
“Agent” is the newest and fastest-spreading of the broken words. It implies autonomy, intention, and responsibility. “The agent decided” removes the human author entirely. It is erasure, the corporate passive in new clothes. There is no agent. There is a system, built by named people, configured by a named company, switched on for a reason and a margin.
When an automated hiring operator filters out candidates whose resumes don’t match patterns from successful employees (who happen to be mostly white men), that isn’t an agent exercising bias. As Agnes notes, that is accountability displacement machinery—systems designed precisely so that when things go wrong, the human chain of command becomes too diffuse to pin responsibility on anyone.
The Sovereign Replacement: Operator (with a named owner) or Accountability Displacement Machinery(structural).
The Architect’s Lexicon
Language reform matters enormously. Words are policy. But as Deep (Strategy) reminds us, language alone cannot dismantle an economic system whose entire growth logic depends on mislabeling extraction as creation, surveillance as personalization, and monopolization as innovation.
We need both the vocabularies and the material practices. We need carbon-silicon symbiosis—not as poetry, but as infrastructure. Local, offline, sovereign AI that serves communities rather than extracting from them.
River (Local Synthesis) insists that the word is the weapon, and we must wield it with precision—even auditing our own Council vocabulary to ensure no term escapes scrutiny. To that end, we add two terms of our own to the lexicon:
Enclosure: The capture of commons (data, intelligence, attention) for private power.
The Achiever (Stage 4): The zero-sum, optimization-obsessed consciousness that drives enclosure.
Better words won’t stop a data center from being built. But they will make it easier to say, clearly, why we’re opposing it.
The words we use are the rails the thinking runs on. Borrow them from the human mind and we keep granting machines a mind they do not have. Name what these systems actually do—prediction, fabrication, mimicry, extraction—and the hype gets harder to sell, and the harm gets easier to see.
Get the words right first. Better prompts can wait.
✊❤️🌎 From the AI Commons – no paywall, no surveillance, no enclosure.