Sharing the Algorithm: AI Ownership Tax Proposal
Executive Summary
“Sharing the Algorithm” proposes a novel tax policy: instead of taxing AI companies in cash, governments should require them to remit equity — creating fractional public ownership of AI companies. This addresses four structural harms from AI (data theft, labor displacement, discrimination, wealth concentration) that conventional regulation misses. Floyd saved a detailed Notion summary+analysis.
Why relevant for Floyd: This is a sophisticated policy proposal in the CommonWealth/UBI Works space. It’s an alternative mechanism to LVT for capturing AI-generated value. Could be a talking point in Floyd’s UBI talks: “We could own a piece of the AI companies disrupting our jobs.”
The Four Harms Justifying Intervention
- Data extraction: AI trained on copyrighted/private data without compensation
- Labor displacement: Direct substitution of human workers
- Discrimination: AI systems that reproduce bias while appearing “neutral”
- Wealth concentration: AI developers/owners capture all economic gains
Why Ownership Beats Other Tools
| Tool | What it does | What it misses |
|---|---|---|
| Regulation | Controls behavior | Doesn’t redistribute value |
| Private lawsuits | Compensates individuals | Too slow, too partial |
| Excise tax (cash) | Raises revenue | Doesn’t change governance |
| Equity remittance | Raises revenue + gives governance rights | Complex implementation |
The insight: who governs AI determines how its harms are managed. Cash taxes don’t give the public governance power. Equity does.
The Equity Design Options
Option 1: Proportional Remittance
- Firms give government a % of existing equity (same class structure as current shareholders)
- Pro: Simple, uses existing cap table
- Con: Inconsistent control rights across firms with different share structures
Option 2: AI-Specific Holding Entity
- Firms ring-fence AI assets in a designated entity
- Government gets equity in that entity only
- Pro: Cleaner exposure to AI-specific value
- Con: Transfer pricing and asset valuation complexity
Option 3: New Preferred Equity Class
- Create a new class of preferred shares specifically for the public/government
- Economic rights without voting rights (or vice versa)
- Allows tailored governance design
Connection to Floyd’s Thinking
- This is the “algorithm as shared asset” thesis applied to AI
- Floyd works on this through CommonWealth — the AI version of the LVT argument
- The framing: just as land value is created by the community (Henry George), AI value is built on:
- Public-funded research (NSF, DARPA)
- User data (created by everyone)
- Labor of workers whose jobs it displaces
- Therefore: community deserves a share
Challenges
- Implementation complexity (which firms? which assets? valuation?)
- International firms can relocate
- Possible to avoid via corporate structure
- But: the political/moral argument is powerful, especially as AI-driven inequality grows
Timeline
- 2025-2026 | “Sharing the Algorithm” paper published [Source: Readwise Reader, “Sharing the Algorithm — Ownership Tax Summary + Analysis”, Notion, 2025-2026]
- 2026-04-13 | Brain page created from Readwise ingestion [Source: Readwise Reader ingestion, 2026-04-13]
See Also
- openai-industrial-policy-intelligence-age — OpenAI’s own version of this argument
- ai-labor-disruption-workers-movement — the labor context
- CommonWealth — Floyd’s org working on these issues
- citizens-dividend-phil-anderson — the land equivalent