Civic Dividend: Decoupling Economic Agency from Wage Necessity Before the Circuit Breaks

Doctrine in one sentence

Rather than defending the wage circuit that AI may be eroding, redesign the institutions that convert productive capacity into mass economic agency, so that households retain purchasing power, democratic legitimacy and meaningful choice regardless of who or what performs cognitive work.

Executive summary

The Unit Cost Dominance thesis is a conditional but serious warning. Its force is not that all jobs vanish by a fixed date, but that if broadly capable AI can perform cognitive work more cheaply than humans at scale, the wage-demand circuit loses its structural role before alternative institutions exist to replace it. The correct policy response is not to inflate substitution costs artificially or to promise that retraining will restore the old equilibrium. It is to build successor institutions now, while the labour market still generates the tax base to fund them, and while democratic coalitions for doing so can still be assembled.

This proposal rests on four pillars. First, an observable-base compute levy whose revenue is ring-fenced for household dividend payments, structured to avoid avoidance and import leakage. Second, a statutory Participation Income that replaces piecemeal means-testing and attaches economic agency to contribution broadly defined, not to wage employment alone. Third, a Sovereign Productive Asset scheme that expands public and quasi-public equity in high-value AI infrastructure, generating non-wage returns that can be distributed. Fourth, a Competition and Portability regime that prevents platform and housing rents from absorbing any income gains households receive. Each pillar has bounded-regret value: each is defensible on ordinary public-policy grounds even if AI disruption is slower or narrower than the thesis suggests.

The package does not require measuring the AI share of any workflow. It does not require global coordination. It is designed to survive hostile future governments through institutional lock-in and automatic stabiliser mechanics. It distinguishes actions the UK can take alone, actions requiring EU-level frameworks, and actions that member states can implement within existing treaties.

The policy package

Pillar 1: Compute Contribution Levy (CCL). A levy applied to commercial purchases of high-performance compute capacity above a de minimis threshold, denominated per accelerator-hour or equivalent floating-point benchmark. The base is observable through data-centre metering, cloud provider invoicing and customs declarations for imported hardware. The rate begins low (indicatively 3 to 5 per cent of commercial compute spend) with a schedule for upward revision tied to independently audited indicators of labour market displacement. Revenue flows into a constitutionally ring-fenced Civic Dividend Fund. The levy applies equally to domestic and foreign-hosted compute used by resident firms, preventing offshoring arbitrage. It interacts with existing corporation tax through partial crediting to avoid cascade, and is scored by the OBR or equivalent before enactment.

Pillar 2: Participation Income (PI). A universal payment conditional on broadly defined social contribution including employment, registered caregiving, approved education and training, volunteering, and civic participation, but not conditional on means or job search alone. PI replaces the household income floor currently provided by an underfunded patchwork of Universal Credit, legacy benefits and tax credits. It is set initially below the level of full minimum-wage employment to preserve work incentives at the margin, with a path to convergence if CCL revenues permit. PI is paid through existing HMRC and DWP infrastructure in the UK, and through member-state transfer systems in the EU. Capital rules are relaxed so that modest savings do not disqualify households.

Pillar 3: Sovereign Productive Asset Scheme (SPAS). Governments use existing public investment vehicles (UK National Wealth Fund, EU InvestEU, member-state development banks) to take equity and convertible positions in AI infrastructure, compute provision and high-value platform businesses operating in their jurisdiction. Returns are channelled into the Civic Dividend Fund rather than general government revenues. Procurement contracts for AI services include equity-warrant conditions where legally permissible, creating a flow of public ownership without requiring new spending authority for each transaction. This is not industrial policy in the traditional sense: it is a mechanism for ensuring that productivity gains from AI generate public returns at scale.

Pillar 4: Competition, Portability and Rent Suppression. CCL revenue and PI are economically inert if platform rents, housing costs and energy costs absorb them. The fourth pillar uses existing competition and digital regulation powers to require data portability, algorithmic interoperability and open-access conditions on dominant platforms; to extend planning and land-value reform to suppress housing rent capture; and to ensure that regulated energy networks do not benefit asymmetrically from AI-driven demand growth at household expense. None of these require new primary legislation in the first instance: they extend existing regulatory mandates.

United Kingdom: first 24 months

Month 1 to 6: Design and consultation. HMRC and HM Treasury commission a joint feasibility study on the Compute Contribution Levy base, working with the Information Commissioner's Office and the AI Safety Institute on metering methodology. The OBR is asked to score three CCL rate scenarios. The Department for Work and Pensions publishes a Green Paper on Participation Income, drawing on the existing literature on participation income pilots in Europe and the verified evidence on Universal Credit conditionality. The National Wealth Fund announces a strategic review of its digital and technology mandate with a view to equity rather than only debt instruments.

Month 6 to 12: Primary legislation introduced. A Finance Bill amendment introduces the CCL at the lowest modelled rate (approximately 3 per cent) with a statutory review mechanism linked to ONS displacement indicators. The Participation Income Green Paper closes and a White Paper is published, proposing PI as a top-up to existing benefits for the first Parliament rather than a wholesale replacement, reducing the fiscal cliff risk. Parliament passes a Statutory Instrument extending the Digital Markets, Competition and Consumers Act framework to mandate data portability and API access for AI-adjacent platforms.

Month 12 to 24: Infrastructure and pilots. The National Wealth Fund makes its first equity-warrant-linked AI infrastructure investments. Two regional PI pilots are established (subject to DWP legal authority) in areas with measurable AI-exposed employment concentration, as identified by DSIT. The Civic Dividend Fund is established as a statutory body with an independent board, ring-fenced revenues and a charter preventing ministerial direction on disbursement. Housing: the existing Leasehold and Freehold Reform Act framework is extended via secondary legislation to cap ground rent indexation, preventing landlords from capturing PI receipts.

European Union and member states: first 24 months

EU level. The European Commission proposes a Directive on AI Economic Transition requiring member states to designate a national Transition Competency Authority and publish biennial assessments of AI-linked labour market change using harmonised Eurostat indicators. The AI Act's existing high-risk workplace provisions are strengthened through implementing guidance to require that affected employers notify works councils and public employment services before AI-linked workforce reductions exceed five per cent of a covered category. The Commission uses its existing competition mandate to issue guidance on interoperability and data portability in AI-adjacent markets, building on the Digital Markets Act framework. The EU does not attempt to create a direct EU-level income entitlement (which would require treaty change and unanimous new own resources) but instead sets minimum standards for member-state transfer floors as a condition of accessing AI investment support under InvestEU.

Member state level. France, Germany, the Netherlands and Sweden, which have existing active labour market institutions and social insurance infrastructure, are the natural early movers. Each can introduce a compute-linked contribution within its own tax system without EU coordination. Works council rights under existing co-determination law in Germany provide a model for AI deployment notification that other member states can replicate. Smaller member states with less institutional capacity are supported through the EU Just Transition Fund, extended to include AI-transition planning as an eligible expenditure. Member states are explicitly permitted (but not required) to link access to AI procurement support to employer commitments on transition notification and retraining.

Years 3 to 5 and dormant triggers

By year three, OBR scoring of the CCL will have produced at least one full revenue cycle. If the fund is generating revenue in line with projections and ONS displacement indicators remain below agreed thresholds, no further action is required: the package operates in maintenance mode.

Trigger 1: Displacement acceleration. If ONS data show that AI-exposed occupations have experienced employment rate declines of more than three percentage points beyond trend over any rolling two-year period, the CCL rate rises automatically to the next pre-legislated band (indicatively 5 to 8 per cent) and the PI level rises to 80 per cent of the National Living Wage equivalent. No further parliamentary vote is required for the rate change: the bands and triggers were set in primary legislation.

Trigger 2: Revenue shortfall. If CCL revenue falls below the Civic Dividend Fund's minimum disbursement floor (to be modelled before enactment), SPAS returns are automatically redirected from reinvestment to disbursement. If both are insufficient, the statutory review board must report to Parliament within 60 days with alternative funding options. The government is required to respond within 90 days: silence does not constitute approval for cuts to PI.

Trigger 3: Concentration. If the Competition and Markets Authority or the European Commission finds that three or fewer entities account for more than 60 per cent of commercial AI compute provision in the jurisdiction, mandatory structural separation or access pricing obligations are triggered without requiring a separate investigation.

Contingent measures held in reserve. A robot or AI output tax (applied to verified autonomous AI outputs above a volume threshold) is designed but not enacted. Its design is published for consultation, it is scored by the OBR, and it enters a legislative queue that can be activated by statutory instrument if Trigger 1 fires and CCL revenues alone are insufficient. This avoids the political cost of enacting it prematurely while ensuring it is not invented under crisis conditions.

Funding and fiscal arithmetic

The honest position is that the arithmetic is incomplete and must be modelled before enactment. What can be stated from the verified facts:

CCL revenue estimate (indicative only). UK commercial AI compute spend is not separately published. The broader digital economy and data-centre sector gives a rough anchor. A 3 per cent levy on a conservatively estimated £8 to £12 billion annual commercial compute base yields £240 to £360 million per year at the initial rate. This is insufficient to fund a PI at meaningful scale alone. It is sufficient to fund pilots, the Civic Dividend Fund administration, and the SPAS equity stake programme in the first Parliament. The Treasury's pre-enactment modelling must provide a central estimate and sensitivity range.

SPAS returns. These are genuinely uncertain and long-lagged. Equity positions in early-stage infrastructure may take five to ten years to generate material returns. The National Wealth Fund's existing £27.8 billion capacity provides a ceiling for the initial equity programme; returns are not a near-term funding source.

PI transition cost. Replacing Universal Credit and legacy benefits with PI would require significant upfront reform costs and could increase the headline transfer bill depending on eligibility breadth. A top-up model in the first Parliament limits the gross cost increase. The net fiscal effect depends on the extent to which PI replaces existing administration costs, conditionality enforcement and sanctions-related expenditure. The OBR must score this before the White Paper becomes a Bill. The proposal cannot honestly claim fiscal neutrality without that scoring.

Conclusion on fiscal credibility. This proposal identifies plausible revenue sources and sequencing but cannot close the arithmetic gap without Treasury modelling. It says so explicitly. A proposal that claimed otherwise would be hiding the gap.

Political coalition and public case

Who benefits. Care workers, the self-employed, gig workers and part-time workers who currently fall outside employer National Insurance coverage gain from PI. Small businesses that use AI to compete with larger incumbents benefit from a compute levy that falls proportionally harder on large-scale commercial users. Pensioners benefit from SPAS returns if linked to the state pension triple lock supplement. Young workers facing a structurally changing labour market gain the insurance value of PI regardless of career trajectory.

Who pays and who loses. Large cloud providers and data-centre operators pay the CCL. Large employers who have substituted AI for cognitive labour pay through a higher effective total cost of AI deployment. Landlords lose some rent-capture capacity under the portability and rent-suppression pillar. These groups will lobby against the package.

Ordinary-language public case. "We built a National Insurance system when most people worked in factories. Most of us now work with data and ideas, and some of that work is being done by machines. We are not going to pretend that is not happening, and we are not going to punish workers for it. We are going to make sure that the gains from automation are shared, that families have a reliable income floor, and that public investment in AI technology earns a return for the public, not just for shareholders."

Sequencing. CCL first, because it raises revenue and signals intent without committing to a welfare architecture that requires detailed design. PI second, starting with pilots and a top-up structure. SPAS third, as equity positions take time to structure. Competition reform runs concurrently throughout.

Compensation for losers. Large compute providers are offered regulatory certainty (a clear, published levy schedule rather than discretionary charges) in exchange for accepting the levy. Employers facing displacement obligations receive partial credits against CCL liability for documented retraining expenditure.

Durability and anti-capture design

Ring-fencing. The Civic Dividend Fund is constituted by primary legislation with a board appointed jointly by Parliament, the devolved governments and an independent nominations committee. Ministerial direction on disbursement is statutorily prohibited. Revenue can only be redirected to non-PI purposes by a two-thirds Parliamentary majority, making raids politically costly.

Anti-avoidance. The CCL base is defined by reference to physical compute infrastructure located in or commercially accessed from the jurisdiction, not by the legal domicile of the provider. Import treatment mirrors the principle of the Carbon Border Adjustment Mechanism: overseas compute used by UK or EU resident firms is assessed on an equivalent basis. This requires bilateral or multilateral information-sharing agreements with major cloud-provider domiciles; these must be negotiated and are a genuine obstacle.

Offshoring. Firms that shift cognitive workloads to overseas AI providers but sell outputs in UK or EU markets are caught by a deemed-access rule applied through the CCL. Enforcement is imperfect for small transactions but covers the bulk of commercial value.

Hostile future governments. The automatic trigger mechanism reduces the discretion available to a future government to quietly defund PI during a fiscal squeeze. Cuts to the PI level below the statutory minimum require an affirmative Commons vote, creating a public record. The SPAS board's independence insulates equity management from short-term political pressure.

Concentration of administrative power. The Civic Dividend Fund board is explicitly prohibited from holding equity, setting procurement policy or advising on levy rates. These functions are separated across HMRC (levy), National Wealth Fund (equity), and the Fund (disbursement). No single institution controls the whole chain.

Legal and institutional obstacles

UK. Universal Credit replacement requires primary legislation and DWP systems reform at substantial cost and risk. The CCL requires careful State aid-equivalent analysis in the context of UK subsidy control rules. The OBR's scoring requirement creates a genuine constraint on timetable. Housing rent suppression through secondary legislation may be challenged as beyond the scope of existing powers: primary legislation may be needed.

EU level. New own resources for an EU-level income transfer require unanimity, making a direct EU PI impossible in the near term. The AI Transition Directive route requires qualified majority voting in most respects but faces political resistance from member states wary of labour market harmonisation. Extending the DMA's interoperability mandate to AI-adjacent markets is legally available but will face legal challenge from affected platform providers. Article 123 TFEU rules out ECB financing of any component of this package: this is a hard constraint and is respected throughout.

Information-sharing for CCL anti-avoidance. Deemed-access assessment for overseas compute requires tax information exchange agreements with the United States, Ireland and Singapore at minimum. These take years to negotiate and may not be achievable in the first Parliament. The levy therefore starts with a domestic-compute base and extends as agreements are concluded.

Failure modes, review and exit rules

Failure mode 1: CCL revenue too small to fund meaningful PI. Review point: end of year two. If CCL revenue at the initial rate is less than 60 per cent of the pre-enactment central estimate, the statutory board must propose either a rate increase, an alternative revenue base, or a reduction in PI scope. Parliament votes within one session.

Failure mode 2: AI disruption does not materialise at scale. Review point: end of year five. If ONS displacement indicators show no statistically significant divergence from pre-AI trends across AI-exposed occupational categories, the CCL rate reverts to its minimum band, PI remains as a modernised Universal Credit replacement (retaining its ordinary public-policy benefits), and SPAS continues as a standard sovereign wealth function. The package does not require disruption to justify its existence at baseline.

Failure mode 3: SPAS equity positions underperform. The NWF's existing convertible and equity instruments already carry performance risk. SPAS underperformance reduces the long-term dividend stream but does not break the CCL-funded PI floor. The Fund's disbursement minimum is met from levy revenue alone; SPAS returns are treated as upside. If a SPAS position becomes a liability, it is managed under existing NWF governance with no recourse to the Civic Dividend Fund.

Failure mode 4: Political coalition collapses before enactment. The package is sequenced so that the CCL (the most technically self-contained element) can be enacted independently. A CCL alone, ring-fenced but without immediate PI disbursement, builds the fund while the political coalition for PI is assembled. This is not the preferred sequencing but it is survivable.

Falsifiability indicators. The package should be reviewed against: (a) the share of AI-exposed occupations showing employment decline above trend; (b) the ratio of CCL revenue to GDP; (c) the Gini coefficient net of PI transfers; (d) the share of households reporting PI as their primary income source; and (e) the concentration ratio in commercial AI compute provision. Annual ONS and Eurostat data provide observable bases for all five.

Feasibility table

Feasibility ratings are based on legal competence, institutional capacity and political coalition analysis as described in the body text. Arithmetic for CCL revenue, PI cost and SPAS returns is explicitly flagged as incomplete pending Treasury and OBR modelling.

What is genuinely new here

Three features of this proposal are not obvious from the thesis diagnosis and deserve separate identification.

First, the automatic trigger mechanism treats the CCL rate and the PI level as linked automatic stabilisers rather than political decisions. This is borrowed from macroeconomic circuit-breaker design, not from AI policy literature. It insulates the package from the normal political economy of austerity, in which transfer programmes are the first to be cut in a fiscal squeeze, precisely when they are most needed.

Second, the deemed-access CCL base, modelled on the Carbon Border Adjustment Mechanism, is designed to close the most obvious avoidance route (shifting compute workloads offshore) without requiring a global coordination agreement. It is imperfect but it changes the calculus for firms considering regulatory arbitrage. No existing AI levy proposal has incorporated this mechanism explicitly.

Third, the package deliberately declines to set a target employment rate or a target for human involvement in AI-augmented workflows. This is a departure from most AI labour policy discussion, which implicitly treats preserving human employment as the success metric. The success metric here is mass economic agency: whether households have reliable income, meaningful choices and democratic voice, regardless of the employment composition of the economy. This reframing is non-obvious and is not derived from the thesis alone.

Bottom line

This package cannot promise that PI will be fully funded from CCL revenues in the first Parliament; the arithmetic requires modelling that does not yet exist, and this proposal says so. What it can promise is that if disruption accelerates, the institutional architecture is already in place to respond at scale without emergency improvisation, and if disruption is slower than feared, every element of the package has ordinary public-policy value that justifies it on conventional grounds. The political coalition is real but not guaranteed: it requires a government willing to frame AI not as a productivity windfall for shareholders but as a civilisational shift that changes the terms on which citizenship is economically expressed. That framing is available. Whether it is taken up is a question of political will, not policy design.