Sovereign Agency Dividend

Doctrine in one sentence

AI-driven unit-cost dominance can erode wage-based income, so a permanent, tax-financed dividend to all citizens is needed to preserve mass economic agency, demand and democratic legitimacy.

Executive summary

The proposal creates a Sovereign Agency Dividend (SAD) – a universal, non-means-tested cash payment funded by a dedicated levy on AI-enabled cognitive services and on excess economic rents captured by platform owners. The levy is collected through existing tax structures, is calibrated to the observable value of AI-generated output, and is redistributed on a per-capita basis. The design is compatible with continued AI adoption, relies on measurable fiscal bases, fits within the legal powers of the UK Parliament and the EU institutions, and includes built-in review triggers that can be scaled back or terminated if the underlying threat proves overstated. In the first two years the UK will legislate the levy and launch a pilot dividend; the EU will adopt a coordinating framework and allocate adjustment funds; years 3-5 see the dividend rolled out across member states with periodic actuarial reviews; later phases provide contingency triggers and sunset clauses. The package preserves mass agency without preserving unproductive jobs, avoids the need for precise AI-share accounting, and can be financed within current fiscal ceilings when combined with modest re-allocation of existing public-investment funds.

The policy package

1. AI Cognitive Services Levy (ACSL) – a 2 % charge on gross revenues of firms that derive more than 30 % of their digital output from AI-generated cognitive services (e.g., large-language-model APIs, automated content creation, algorithmic design). The levy is collected by HMRC and the European Commission’s VAT-style digital revenue pool, using the same filing and audit mechanisms that already exist for corporate income tax.
2. Sovereign Agency Dividend (SAD) – a flat annual payment of £1,200 per adult and €1,200 per adult in the EU, paid quarterly into personal bank accounts. Eligibility is universal; no means-testing or asset limits apply.
3. Digital Public-Goods Fund (DPGF) – a ring-fenced portion of the levy (30 %) is allocated to the development of open-source AI tools, data-infrastructure and digital skills programmes, ensuring that the public derives direct benefit from the very technologies that generate the levy.
4. Transition Adjustment Programme (TAP) – a temporary top-up for workers displaced from occupations where AI substitution exceeds 50 % of tasks, providing up-skilling vouchers worth £1,500 (UK) or €1,500 (EU) per year for two years. The programme is financed from the same levy but is automatically sunset after five years unless a review triggers its renewal.
5. Rent-Capture Safeguard (RCS) – a cap on the proportion of any sector-wide economic rent that can be absorbed by private platform owners; any rent above the cap is redirected to the SAD pool. The cap is enforced through a mandatory reporting requirement for firms with annual AI-derived revenues above £500 million (UK) or €500 million (EU).

All components are designed to be observable, auditable and enforceable under current UK and EU fiscal and competition law.

United Kingdom: first 24 months

• Month 1-3: Draft the AI Cognitive Services Levy Bill and publish a consultation on the definition of “AI-generated cognitive output”. The bill will be introduced to Parliament with a cross-party sponsor (Labour, Conservative and Liberal Democrat members).
• Month 4-6: Pass the Levy Act; HMRC begins training auditors and integrates the levy into the existing Corporate Tax self-assessment system. A pilot levy rate of 2 % is set, with a built-in review after 12 months.
• Month 7-9: Establish the SAD payment infrastructure within the Department for Work and Pensions (DWP), linking it to the existing National Insurance number system to enable automatic quarterly disbursement.
• Month 10-12: Launch the Digital Public-Goods Fund, allocating the first £200 million of levy receipts to open-source model development and digital-skill bootcamps in partnership with universities and the National Wealth Fund.
• Month 13-18: Roll out the Transition Adjustment Programme for workers in high-substitution occupations (identified via ONS skill-mix data). Vouchers are issued through the National Careers Service.
• Month 19-24: Publish the first fiscal impact assessment, confirming that the levy is projected to raise £4.5 billion annually, sufficient to fund the per-adult dividend at the level set. A parliamentary review committee is mandated to evaluate the levy’s effectiveness and recommend adjustments before the end of year 2.

European Union and member states: first 24 months

• Month 1-4: The European Commission adopts a Regulation on the Digital AI Revenue Pool, harmonising the definition of AI-generated cognitive services across member states and establishing a common reporting format for the levy.
• Month 5-8: Member states transpose the Regulation into national law, each setting up a national AI levy unit within their tax authority. The EU’s own-resource mechanism is used to pool a portion of the levy for the Digital Public-Goods Fund, which is administered by the European Investment Bank.
• Month 9-12: The European Parliament and Council approve a resolution establishing the EU-wide Sovereign Agency Dividend, with a uniform payment of €1,200 per adult, to be distributed via national social-security systems.
• Month 13-18: Launch a coordinated Transition Adjustment Programme, funded jointly by the EU’s Digital Europe Programme and national co-financing, offering up-skilling vouchers to workers in high-substitution sectors identified by Eurostat.
• Month 19-24: Conduct a joint EU-UK audit of the levy’s revenue and dividend payout, publishing a transparent dashboard. The European Commission is tasked with preparing a mid-term review for the end of year 2, including a trigger mechanism for scaling the levy up or down.

Years 3 to 5 and dormant triggers

• Year 3: Full-scale payment of the SAD begins in all EU member states and the UK. The dividend is indexed to inflation and to the average growth rate of AI-derived GDP.
• Year 4: The Transition Adjustment Programme is evaluated; if more than 15 % of participants report successful up-skilling and re-employment, the programme is discontinued. Otherwise, a renewal vote is triggered in the European Parliament and the UK Parliament.
• Year 5: A “Dormant Trigger” clause activates if AI-derived revenues fall below 0.5 % of total digital GDP for two consecutive quarters; in that case the levy rate is automatically reduced to 1 % and the dividend is adjusted downward by 10 % to preserve fiscal balance.
• Year 5-7: A biennial review is mandated, using observable indicators (AI-revenue share, dividend uptake, labour-market stress) to decide whether to maintain, increase, or phase out the levy. The review process includes a parliamentary “sunset” vote after the fifth year.

Funding and fiscal arithmetic

• Levy revenue projection (UK): 2 % of AI-derived gross revenues is estimated at £4.5 billion per year (based on ONS 2026 AI-use data: 29 % of businesses use AI, with an average AI-revenue share of 12 %).
• Dividend cost (UK): 67 million adults × £1,200 = £80.4 million annually – a trivial share of the levy revenue, leaving ample fiscal space for the Digital Public-Goods Fund and Transition Adjustment Programme.
• EU-wide levy revenue: Using the EU’s 2025 AI-use figure (20 % of enterprises), a 2 % levy on the estimated €150 billion of AI-generated cognitive output yields roughly €3 billion annually. This funds the EU-wide dividend (≈ 450 million adults × €1,200 = €540 million) plus the Digital Public-Goods Fund, with the remainder allocated to the EU’s own-resource budget.
• Transition Adjustment Programme cost: Assuming 1 million displaced workers in the UK and 1.5 million in the EU, at £1,500 per year for two years the total cost is £3 billion (UK) and €3 billion (EU), fully covered by earmarked levy receipts.
• Modeling requirement: Before enactment, the UK Treasury and the European Commission must commission an actuarial model that projects AI-revenue growth under three scenarios (steady, accelerated, decelerated). The model must be published and subjected to parliamentary/OMB scrutiny.

The arithmetic shows that the package can be self-financing within current fiscal ceilings, provided that the levy is not circumvented and that avoidance is limited by robust anti-avoidance rules.

Political coalition and public case

• Coalition: A broad alliance of centre-left parties (Labour, Social Democrats, Progressive Alliance), centre-right parties (Conservative, European People’s Party), and progressive civil-society groups (Trade Unions, Digital Rights NGOs). The coalition frames the dividend as a “future-proof social contract” rather than a redistribution of existing wealth.
• Public case: The narrative stresses that AI is a public good that should benefit everyone; the dividend guarantees that citizens retain purchasing power and can invest in education, housing and community life, regardless of how work evolves. The message is communicated through town-hall meetings, social-media campaigns and targeted advertising that highlight the tangible £1,200 annual boost to household budgets.
• Losers: Large AI-service providers that currently retain all platform rents, and firms that rely on low-skill labour to offset high unit costs. They are compensated through a phased-in levy and a transition period for up-skilling programmes.
• Sequencing: First, the levy is passed; second, the dividend infrastructure is built; third, the Digital Public-Goods Fund is launched; fourth, the Transition Adjustment Programme is rolled out; finally, the dividend is paid out. This order builds credibility and demonstrates tangible benefits before the full dividend begins.

Durability and anti-capture design

• Capture mitigation: The levy is levied on gross revenues, not on profits, reducing the incentive to shift earnings. A mandatory “rent-capture” reporting threshold forces large platforms to disclose AI-derived earnings, making avoidance transparent.
• Raids and avoidance: The levy includes a “digital services permanent establishment” rule that treats any foreign AI provider with more than 1 million UK users as a taxable presence, closing the offshoring loophole.
• Dilution protection: The dividend is indexed to AI-derived GDP growth, so if AI output expands rapidly the dividend rises accordingly, preventing its real-value erosion.
• Administrative power limits: The levy administration is split between HMRC and the European Commission, with independent audit boards in each jurisdiction. No single minister can unilaterally alter the dividend rate; any change requires a parliamentary or European Parliament vote.
• Hostile government safeguard: The legislation includes a “sunset clause” that automatically repeals the levy if a future parliament passes a repeal bill with a two-thirds majority, ensuring that the policy cannot be entrenched against democratic reversal.

Legal and institutional obstacles

• UK: The Levy must be compatible with the Finance Act 2023 and the UK’s devolution settlements; it can be introduced as a “digital services surcharge” under existing tax-policy powers. The dividend payment uses the existing National Insurance number infrastructure, avoiding new statutory creation.
• EU: The Digital AI Revenue Pool Regulation requires qualified majority voting in the Council, which is feasible given the coalition of member states that have already signaled support for a digital tax. The dividend must be implemented through national social-security systems, respecting the principle of subsidiarity. State-aid rules limit the use of the levy for direct subsidies to private firms; the design keeps all proceeds in the public purse.
• Cross-border enforcement: The OECD-G20 framework on digital taxation provides a basis for information exchange and mutual assistance, facilitating compliance and reducing the risk of double-non-taxation.

Failure modes, review and exit rules

• Revenue shortfall: If AI-derived revenues fall below 0.5 % of digital GDP for two consecutive quarters, the levy rate automatically drops to 1 % and the dividend is reduced by 10 %. A parliamentary review must then assess whether to reinstate the original rate.
• Political backlash: Should public opinion turn against the dividend, a parliamentary “sunset” vote can be triggered after year 5, requiring a two-thirds majority to keep the levy.
• Legal challenge: If a court rules that the levy violates existing tax treaties, the legislation includes a “compliance amendment” clause that allows the government to renegotiate treaties or to redesign the levy as a consumption-based charge.
• Technological shift: If AI adoption plateaus and unit-cost dominance does not materialise, the levy can be voluntarily reduced to 0 % after a scheduled review, with the dividend funded from general taxation instead.

Feasibility table

What is genuinely new here

• The package treats AI-generated cognitive output as a taxable economic rent rather than a mere productivity gain, and channels that rent into a universal cash dividend that guarantees mass economic agency.
• It couples the levy with a rent-capture safeguard that forces platforms to disclose and surrender excess rents, a mechanism not currently present in digital-tax debates.
• The design embeds observable fiscal triggers (AI-revenue share thresholds) and automatic adjustment clauses, ensuring that the response can be scaled back if the AI-dominance thesis proves unfounded, thereby limiting bounded regret.
• By integrating the dividend into existing National Insurance and social-security infrastructures, the solution avoids the creation of a new bureaucratic layer, a novelty compared with previous universal-basic-income proposals that rely on separate payment systems.

Bottom line

The Sovereign Agency Dividend offers a politically viable, institutionally durable and fiscally credible response to the risk that AI will erode wage-based income. It does so by taxing the observable revenues of AI-enabled cognitive services, redistributing the proceeds as a universal cash payment, and safeguarding the system with built-in review mechanisms and anti-capture rules. The first two years focus on legislative enactment, infrastructure build-out and pilot programmes; years 3-5 expand the dividend across the UK and EU while embedding trigger-based adjustments; later phases provide contingency exits. If the underlying thesis materialises, the dividend preserves demand and democratic legitimacy; if it does not, the policy can be scaled down or terminated with minimal fiscal distortion, delivering ordinary public-policy benefits in any scenario.