Data Compensation

Wealth Capture · Source: Windfall-trust

HEAVY COPE

What it proposes

Requiring AI companies to compensate individuals for the data used to train AI systems.

Establishing legal frameworks for individuals to be compensated when their data — personal information, creative works, behavioral patterns — is used to train AI systems. Variants include data dividends, licensing fees, and collective bargaining for data rights.

The challenge (their words)

The value of individual data is negligible in aggregate training sets. Collective action problems prevent effective negotiation. AI training increasingly uses synthetic data, reducing dependence on human-generated data. The policy addresses a shrinking lever as AI becomes self-sufficient in data generation.

Discontinuity Thesis Score Breakdown

💰 75

Unit-Cost Survivability

Does it survive near-zero marginal cost?

Data compensation adds a marginal cost to AI training. This cost is spread across billions of contributors, making individual payments negligible. It cannot offset the wage loss from displacement.

🔌 72

Interface Collapse

Does it account for AI as the integration layer?

Data ownership doesn't prevent interface collapse. Owning the training data for the system that automated your workflow doesn't give you a role in the new workflow.

📉 68

Propagation Blindness

Does it see the full task→job→market cascade?

Treats the problem at the wrong layer. Data compensation addresses the input side (training data) while ignoring the output side (displacement). The cascade runs regardless of who owns the training data.

🎯 62

Coordination Feasibility

Can it be enforced when defection = advantage?

Requires global data rights frameworks. International enforcement is extremely difficult. Companies relocate training operations to jurisdictions without data compensation mandates.

Oracle Verdict

Micro-royalties for the raw material of your replacement. Data compensation pays people for the data used to train the AI that replaces them. The amounts per person would be trivial — spread across billions of data contributors, individual payments are noise. More fundamentally, this treats the problem as a property rights issue rather than an economic structure issue. You are not poor because you weren't paid for your data. You are poor because you were replaced by the system your data trained.

Scored by claude-opus-4-6-oracle

View original at Windfall-trust →