Sabotage (Veblen)

In the AI Story

Veblen introduced the concept in The Engineers and the Price System (1921), arguing that productive capacity of the industrial system was systematically underutilized not because of technical limitation but economic calculation. The business enterprise, organized around profit rather than production, found unrestricted output drove prices below levels sustaining profit. The remedy was restriction — what Veblen called 'conscientious withdrawal of efficiency' — a deliberate choice to produce less than the system could produce for maintaining pecuniary returns the system was organized to generate.

AI sabotage operates through tiering of capability. Large language models aren't offered as single undifferentiated services but in tiers — each providing different capability levels at different prices. Basic tiers provide models adequate for casual use. Professional tiers provide more capable models or higher usage limits. Enterprise tiers provide capabilities (advanced reasoning, extended context, specialized tools) withheld from lower tiers. The tiering is described as pricing structure matching cost to value received. Structurally, it's Veblen's sabotage: deliberate restriction of capability the state of the industrial arts could provide to all, maintaining price differentials upon which platform profit depends.

A second form operates through designed dependency. AI tools are designed — not maliciously but structurally, as consequence of business incentives — to create dependency rather than independence. The ideal outcome from the platform's perspective is users becoming progressively more dependent on tools — workflows increasingly organized around tool capabilities, data increasingly stored within tool ecosystems, skills increasingly adapted to tool interfaces rather than underlying domains. The dependent user is recurring revenue. The independent user who acquires capability and exercises it without the tool is a customer lost. Dependency is cultivated through design choices individually defensible as user-experience improvements but collectively a retention system.

A third form — the most structurally significant — operates through training data appropriation. Models train on intellectual output of the community — books, code, articles, creative works produced by millions over decades. Training data is part of the state of the industrial arts. Models ingest this inheritance, process it into commercially valuable products, offer products back to communities at prices. The mechanism is structurally identical to eighteenth-century agricultural commons enclosure. Commons was appropriated by private interests, fenced off, converted into private income sources. Communities that contributed to commons through generations of collective use were required, after enclosure, to purchase access to resources they collectively produced.

Origin

Veblen developed sabotage as a technical concept in The Engineers and the Price System (1921), though the observation runs through his earlier work. The concept emerged from his recognition that business prosperity often required restricting production — a paradox classical economics couldn't explain. If the goal of economic activity is satisfying human wants, why would rational actors deliberately produce less than they could? Veblen's answer: because the goal isn't satisfying wants but generating profit, and profit often requires scarcity.

The concept was controversial and remains so. Veblen's critics argue he overestimated business malice and underestimated genuine efficiencies of market coordination. His defenders note he documented specific, observable instances of deliberate restriction across industries. The AI application is less controversial in mechanism (tiering, dependency design, commons appropriation are observable) but fiercely contested in their characterization as 'sabotage' versus 'normal business practice' — a debate that precisely reproduces Veblen's original point about how institutional vocabularies normalize predatory extraction.

Key Ideas

Restriction, not destruction. Sabotage is deliberate curtailment of output to maintain profitable scarcity, not malicious damage to productive infrastructure or capability.

Normal price-system function. Not aberration but predictable behavior of enterprises organized around profit generation rather than output maximization.

Capability tiering as restriction. AI platforms deliberately withhold available capabilities from lower subscription tiers to maintain upgrade incentives and price differentials.

Designed dependency as enclosure. Tools structured to make users progressively dependent convert independence (one-time capability acquisition) into recurring revenue (ongoing subscription necessity).

Training data appropriation. Using community-produced intellectual output to train proprietary models constitutes enclosure of commons — collective inheritance converted to private property offered back at price.