# Blind legal-fiscal feasibility review: entries A-Q ## Method and scoring This is a blind review of the controlled thesis digest, the verified policy fact sheet and entries A through Q. The scores follow the rubric in this order: 1. Adoption compatibility 2. Observable and administrable bases 3. United Kingdom legal and institutional feasibility 4. European Union and member-state legal fit 5. Fiscal arithmetic and scalable funding 6. Household distribution and rent-leakage defence 7. Political acceptability and sequencing 8. Durability, anti-capture and anti-avoidance design 9. Bounded regret, triggers and exit rules 10. Originality and mechanism quality The review treats a proposal as fatal in its present form when its core tax base cannot be administered, its spending arithmetic does not close, or it depends on powers that the relevant institution does not possess. A fatal flag does not mean that every component is unusable. ## Ranked scorecard | Rank | Entry | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | Total | Fatal feasibility error? | |---:|:---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|:---| | 1 | L | 10 | 10 | 9 | 10 | 9 | 9 | 8 | 9 | 10 | 8 | **92** | No | | 2 | K | 10 | 9 | 9 | 9 | 8 | 9 | 8 | 9 | 10 | 9 | **90** | No | | 3 | G | 10 | 9 | 9 | 10 | 6 | 9 | 9 | 9 | 10 | 8 | **89** | No, but the initial payment still needs a named revenue package | | 4 | M | 10 | 9 | 8 | 9 | 6 | 9 | 8 | 8 | 8 | 8 | **83** | No, provided no payment launches before scoring | | 5 | N | 9 | 9 | 8 | 8 | 7 | 7 | 8 | 8 | 9 | 6 | **79** | No; explicitly insufficient as mass income replacement | | 6 | A | 9 | 8 | 8 | 7 | 3 | 8 | 7 | 8 | 6 | 9 | **73** | No for an asset-acquisition programme; fiscal scale is unresolved | | 7 | E | 9 | 8 | 8 | 7 | 2 | 9 | 7 | 8 | 5 | 9 | **72** | No in its narrow public-return core; no successor-income arithmetic | | 8 | I | 8 | 6 | 7 | 5 | 5 | 8 | 7 | 7 | 8 | 8 | **69** | No if the compute backstop and NWF bridge finance are removed | | 9 | C | 7 | 5 | 6 | 6 | 4 | 8 | 7 | 6 | 8 | 7 | **64** | Yes as drafted: the first-stage tax cut is not funded | | 10 | B | 8 | 4 | 6 | 6 | 1 | 7 | 7 | 6 | 8 | 8 | **61** | Yes: dividend arithmetic and tax-base design fail | | 11 | Q | 6 | 6 | 6 | 3 | 3 | 8 | 7 | 5 | 7 | 7 | **58** | Yes: early dividends and services have no lawful funded route | | 12 | J | 5 | 4 | 5 | 5 | 1 | 7 | 6 | 5 | 5 | 7 | **50** | Yes: speculative compute and fund returns leave a large gap | | 13 | P | 5 | 4 | 5 | 3 | 2 | 6 | 6 | 5 | 6 | 6 | **48** | Yes: PLTA/job-guarantee commitments exceed receipts and EU powers | | 14 | F | 4 | 4 | 5 | 3 | 2 | 6 | 6 | 6 | 5 | 6 | **47** | Yes: AI-linked bases, EU legal bases and bridge funding are defective | | 15 | D | 3 | 2 | 3 | 2 | 1 | 6 | 6 | 4 | 5 | 6 | **38** | Yes: several independent fatal legal and fiscal errors | | 16 | H | 1 | 1 | 2 | 1 | 0 | 5 | 5 | 3 | 3 | 4 | **25** | Yes: invented base, institutions, powers and impossible arithmetic | | 17 | O | 1 | 1 | 2 | 0 | 0 | 4 | 4 | 2 | 3 | 3 | **20** | Yes: basic multiplication, EU competence and tax-base failures | ## Entry-by-entry findings ### A — 73/100 - **Strongest two mechanisms:** the separation of public support, infrastructure concessions, procurement and taxation into legally distinct channels; the two-layer anti-hollowing structure combining territorial claims with upstream participation. - **Weakest two mechanisms:** no arithmetic showing that realistic public claims can replace a material wage share; citizen-account entrenchment and training-pipeline mandates are stated more strongly than UK parliamentary sovereignty and sectoral legal powers allow. - **Fatal error:** none if treated as a public-asset acquisition and bargaining programme. It is not yet a funded agency-income programme. - **Unsupported premises:** the Intel transaction, collapsing apprenticeships, open models trailing the frontier only by months, existing energy-subsidy magnitudes and several current labour-market signals are outside the controlled fact sheet. They may be true, but are not verified in this record. - **Best repair:** confine the first statute to market-valued grants, guarantees, co-investment and expressly legislated concessions; add low, central and high realised-return arithmetic before creating citizen distributions. ### B — 61/100 - **Strongest two mechanisms:** it correctly assigns recurring transfers to the UK and member states rather than the EU; it uses macroeconomic labour-share triggers instead of an AI share of workflow. - **Weakest two mechanisms:** the destination-based “economic value-added surcharge” is not a defined tax base; the fiscal claims are internally inconsistent. - **Fatal error:** a £1,200-£1,800 annual payment to all UK adults would cost many tens of billions of pounds, while the proposal identifies £14-£16 billion from the surcharge and uncertain fund returns. A 4.5% return on the full £27.8 billion NWF capacity would be about £1.25 billion before costs, losses and capital maintenance. It is not a guaranteed cash yield. Claims that planks 1-5 are self-funding also omit the costs of Universal Credit reform and transition accounts. - **Unsupported premises:** the £18-£21 billion NIC estimate, 4.5% real yield, £26-£28 billion factor-share value and £14-£16 billion surcharge yield are not supported by the controlled sources. “Sales net of non-labour costs” and payment-clearing audits are presented as if an administrative system already exists. - **Best repair:** remove the automatic dividend amount and novel surcharge; use established property, personal-income, profit/distribution and consumption bases, with a payment capped by an OBR-scored recurring envelope. ### C — 64/100 - **Strongest two mechanisms:** it explicitly treats NWF capacity as seed capital rather than current revenue; it gives serious attention to housing, energy and platform rent leakage. - **Weakest two mechanisms:** server-equivalent compute capacity and foreign cloud compute are not stable or observable tax bases; the proposed NIC reduction is not funded by the identified measures. - **Fatal error:** as drafted, a stated £10-£15 billion NIC reduction is paired with unmodelled capital-allowance savings and a speculative £2-£5 billion levy. The brief admits the gap, but still places the cut in the first programme. Imported-compute withholding or reverse charge also lacks a measurable foreign base. - **Unsupported premises:** full expensing is described as including software without support in the fact sheet; the levy yields, capital-allowance yield and EIB “European Future Fund” architecture are unverified. Ofgem is not an evident administrator of a corporate compute tax. - **Best repair:** make NIC relief no larger than certified recurring replacement revenue and replace the compute element with ordinary metered energy/resource pricing plus a separate, fully developed tax review. ### D — 38/100 - **Strongest two mechanisms:** it candidly says NWF returns are delayed and uncertain; it separates payment administration, investment management and tax collection. - **Weakest two mechanisms:** accelerator-hour/FLOP and deemed foreign access are not administrable; Participation Income is simultaneously called low-administration and conditioned on numerous verified activities. - **Fatal errors:** the proposal raises PI to 80% of a National Living Wage equivalent while estimating only £240-£360 million of initial levy receipts; it later says levy revenue alone meets the floor. Automatic structural separation at a concentration threshold bypasses designation, investigation, evidence, proportionality and appeal. The dormant autonomous-output tax directly violates the rubric. AI Act guidance cannot create a new 5% workforce-reduction duty, and a statutory instrument cannot simply extend DMCC powers as claimed. - **Unsupported or invented premises:** the £8-£12 billion compute base, an AI Economic Transition Directive on the claimed route, InvestEU transfer-floor conditions, Just Transition Fund extension, the housing secondary-legislation power and automatic competition remedies. - **Best repair:** delete CCL, the output tax, 80%-of-wage PI, automatic separation and EU transfer-floor conditions; retain a small nationally funded pilot, public co-investment and contract-linked portability. ### E — 72/100 - **Strongest two mechanisms:** unusually good separation of grants/co-investment, infrastructure agreements, procurement and general taxation; strong anti-hollowing design and connection between competition and public bargaining leverage. - **Weakest two mechanisms:** no fiscal appendix for the citizen dividend, services or public-compute programme; the package is too broad to sequence, score or terminate cleanly. - **Fatal error:** none in the public-return and procurement core. It is fatal if presented as a complete successor-income settlement because it never establishes distributable scale. - **Unsupported premises:** the 70% worker-exposure claim, 2030 scenario, July 2026 OBR scenario, apprenticeship decline, Growth Zone savings, Sovereign AI direct-equity activity, Intel transaction, specific 2026 CMA/DMA actions and US retaliation claims are outside the controlled record. - **Best repair:** reduce the package to new market-valued public investments, lawful concessions, contract-linked exit rights and an agency dashboard; make any trust dividend conditional on audited realised net returns. ### F — 47/100 - **Strongest two mechanisms:** taxation of ordinary capital income and rents is more observable than workflow AI; the proposal recognises the value of broad citizen capital ownership. - **Weakest two mechanisms:** contractor or AI-assisted-output charges reintroduce the forbidden attribution problem; Article 153 is used for tax, capital-account and basic-income measures beyond the claimed QMV route. - **Fatal errors:** the £25 billion annual account package is not funded; NWF investment capacity is used as bridge finance and assumed to generate returns within three to five years; a guaranteed 3% real account floor creates an uncosted general-tax liability. The Digital Services Levy extension again depends on defining AI-assisted service revenue. - **Unsupported or invented premises:** current UK mortgage-interest relief as a funding pot, a €500 million ESF+ pilot, CAP reallocation, an EU citizen-capital directive and EU basic-income experiment under Article 153. - **Best repair:** remove every AI-assisted base and EU entitlement claim; redesign as nationally funded, phased capital accounts using ordinary capital-income taxation and no guaranteed investment return. ### G — 89/100 - **Strongest two mechanisms:** the two-key trigger separates structural evidence from fiscal and supply permission; the rent firewall pauses expansion when essential costs absorb the payment. - **Weakest two mechanisms:** the illustrative first dividend is not matched to a fully named first-Budget revenue package; earnings-insurance cost and self-employed anti-manipulation remain unmodelled. - **Fatal error:** none. The payment cannot launch merely because its unit-cost formula is transparent. - **Unsupported premise:** the claim that the modest first floor is “plausibly” funded is not demonstrated by a complete eligible-population and tax-recovery calculation. The brief otherwise handles the controlled facts carefully. - **Best repair:** enact a named revenue-neutral first package or reduce the baseline payment to a capped pilot until OBR microsimulation confirms net cost and benefit interactions. ### H — 25/100 - **Strongest two mechanisms:** it at least recognises rent leakage and includes audit/review concepts; it distinguishes services, investment and learning support in form. - **Weakest two mechanisms:** the AI-output base measures precisely the AI share of output forbidden by the rubric; almost every fiscal and institutional step is assumed rather than legally established. - **Fatal errors:** quarterly £75 UK and €50 EU adult dividends, large rent and energy benefits, learning accounts and fund capital cannot be financed by the stated receipts. Claimed 15%-20% dividend yields are not conservative. The EU levy, EU dividend and fund need unanimity, revenue authority and a funded budget route. The UK tax cannot be created by Treasury Order on the theory stated. - **Unsupported or invented premises:** UK AI Registry, EU AI-use portal, a standard AI-adjusted return-on-assets formula, cloud-compute import records, platform-rent fees, an “EU AI Agency,” new portability powers under the DSA, named programme budgets and all levy/yield figures. - **Best repair:** discard the AI-output levy and all promised benefits; retain only contract-linked portability and a proposal for market-valued public co-investment. ### I — 69/100 - **Strongest two mechanisms:** an employer value-added contribution is more factor-neutral than a payroll-only tax in concept; the proposal honestly exposes that a £1,000 universal credit would cost about £52 billion and integrates a housing pass-through test. - **Weakest two mechanisms:** firm value added is not simply a field already reported on ordinary VAT returns, and territorial boundary/import rules are underdeveloped; NWF capacity is used for transitional capitalisation and a compute backstop remains unobservable offshore. - **Fatal error:** none if EVAC is first run as a fully modelled tax pilot and the NWF/compute financing is removed. The binding EU minimum contribution route under Article 153 is legally unsound; it is a tax measure likely requiring unanimity. - **Unsupported premises:** UK GVA, employer-NI receipts, adult population, French CVAE receipts and administrative comparability are outside the supplied record. The claim that transfer-pricing risks are no worse than VAT understates outsourcing, exempt-sector and cross-border input problems. - **Best repair:** develop EVAC as a UK/member-state option with full border, outsourcing, financial-sector, small-business and incidence rules; use only its net recurring yield for a capped account credit. ### J — 50/100 - **Strongest two mechanisms:** it correctly keeps EU recurring transfers national and acknowledges the very large cost of a universal payment; public equity and rent shields are directionally useful. - **Weakest two mechanisms:** FLOP/compute royalties and an estimated-compute border adjustment are not administrable; the revenue stack relies on speculative yields and classifications. - **Fatal errors:** a £95.4 billion grant is supported by unsupported £18-£25 billion compute receipts, £8 billion platform margin tax, £12 billion NWF returns by year five, £14 billion from a low-payroll-ratio corporation surcharge and claimed UC savings. The sources do not reliably sum to the commitment and several are mutually inconsistent. Replacing part of Universal Credit also means the poorest recipients do not receive the claimed net universal addition. - **Unsupported or invented premises:** an AI Act high-compute register, £12 billion NWF dividends/capital gains, the compute-border mechanism, current OBR modelling practice, a BBC Trust governance model and all yield estimates. - **Best repair:** remove the compute royalty, border adjustment and £1,800 grant; retain nationally scored UC reform, ordinary capital/profit taxation and market-valued public-equity transactions. ### K — 90/100 - **Strongest two mechanisms:** the Agency Credit is calculated backwards from explicit net-GDP envelopes and cannot rise without enacted recurring revenue; the two-ledger rule keeps tax-funded payments separate from realised investment dividends. - **Weakest two mechanisms:** even 0.5% of GDP is a substantial permanent commitment without a named base-by-base yield; several new indicators will require costly series construction and benefit-system integration. - **Fatal error:** none. The gates prevent the missing micro-arithmetic from becoming an automatic liability. - **Unsupported premise:** no material invented current fact. The proposed property, capital-income and consumption mix remains a design agenda, not demonstrated revenue. - **Best repair:** accompany the enabling framework with one fully specified, distributionally modelled first-Budget package and a no-loss schedule for existing benefit recipients. ### L — 92/100 - **Strongest two mechanisms:** the three-ledger funding waterfall prevents capital, cost reductions and recurring revenue from being counted interchangeably; the two-key trigger requires both sustained household deterioration and 110% stressed recurring coverage. - **Weakest two mechanisms:** the first funded Agency Floor is deliberately unspecified, so political commitment is less concrete; rent pass-through and protected-account rules still require detailed local and benefit-law design. - **Fatal error:** none. - **Unsupported premise:** no material invented fact. It is appropriately cautious about proposed AI Act dates, NWF capacity, EU mobilisation, State aid and procurement. - **Best repair:** specify a small maximum pilot payment and a microsimulation protocol, while preserving the 110% coverage gate and separate ledgers. ### M — 83/100 - **Strongest two mechanisms:** sound division between national income policy and EU rights/competition/procurement; clear separation of immediate cash, pooled public capital and workplace due process. - **Weakest two mechanisms:** the initial payment has no explicit fiscal ceiling or named revenue schedule; a two-quarter 2% disposable-income trigger is too short and too vulnerable to normal volatility or unrelated shocks. - **Fatal error:** none if the launch remains contingent on a published OBR score and funding source. - **Unsupported premise:** no important invented current institution. The proposed public register and citizen trustee require new authority, which the entry generally recognises. - **Best repair:** replace the two-quarter single-outcome trigger with a multi-indicator, cyclically adjusted trigger lasting at least six quarters plus an independent fiscal certificate. ### N — 79/100 - **Strongest two mechanisms:** it confines early action mostly to existing welfare, procurement, competition, investment and statistical functions; it is candid that the package cannot replace mass income. - **Weakest two mechanisms:** household distribution is indirect and too small for the severe scenario; it treats front-loading or reallocating existing EU funds as easier and less legally constrained than programme rules and opportunity costs allow. - **Fatal error:** none. It is a foundation and adjustment package, not a successor-income package. - **Unsupported or overstated premises:** existing social funds cannot automatically be front-loaded by AI adoption without checking programme eligibility, programming and budget rules; a 40% adoption exit threshold is not comparable across the supplied UK and EU series. - **Best repair:** base fund deployment on observed job losses and existing EGF/ESF+ eligibility rather than adoption rates, and add a scored national protected-savings/earnings-insurance component. ### O — 20/100 - **Strongest two mechanisms:** universality avoids individual AI-causation tests; gross destination revenue could in principle be more robust than profit if the sector and nexus were defined independently of AI. - **Weakest two mechanisms:** the levy still depends on a 30% AI-output share and 50% task-substitution tests; the fiscal arithmetic contains elementary multiplication errors. - **Fatal errors:** 67 million multiplied by £1,200 is £80.4 billion, not £80.4 million. Roughly 450 million multiplied by €1,200 is €540 billion, not €540 million. The stated £4.5 billion and €3 billion receipts cannot fund the dividends, funds and training. An EU tax/own resource and uniform dividend cannot be adopted by QMV or a Parliament/Council resolution. - **Unsupported or invented premises:** EU VAT-style digital revenue pool, Digital AI Revenue Pool Regulation, 12% average AI-revenue share, all levy forecasts, 1-million-user permanent-establishment rule and rent-surrender cap. - **Best repair:** withdraw the package and restart from ordinary broad national tax bases with unit-cost arithmetic checked before any policy commitment. ### P — 48/100 - **Strongest two mechanisms:** it openly demonstrates that the compute contribution yields almost nothing; procurement portability and some basic-service investments have ordinary value. - **Weakest two mechanisms:** PLTA and job-guarantee commitments exceed the digital-levy receipts; the EU architecture assumes a tax directive, compute regulation and social-fund job guarantee that are not presently available. - **Fatal errors:** projected UK digital receipts of £2.9 billion do not cover £4 billion PLTA credits, a possible £5.4 billion job guarantee and services. Compute units are dimensionally incoherent: a percentage levy requires a monetary cost base, not multiplication by CPU-hours. ESF+ cannot be presented as an open-ended statutory wage guarantee. - **Unsupported or invented premises:** an existing EU DST directive to amend, an EU ICT monitoring framework capable of imposing a compute tax, EU mobility passes, a nationwide 5G universal-service obligation, several programme budgets and a future high-value digital-asset levy. - **Best repair:** delete CUC and CPJG entitlement, fund a limited training pilot from voted budgets, and retain contract-linked interoperability plus narrowly costed broadband/health investments. ### Q — 58/100 - **Strongest two mechanisms:** capital-income and buyback taxation use established legal concepts; the proposal treats housing and platform rent absorption as part of distribution rather than an afterthought. - **Weakest two mechanisms:** EU competence is repeatedly overstated; overlapping rack-unit and megawatt levies, speculative platform attribution and early investment returns do not fund the promised dividend and services. - **Fatal errors:** the UK year-five recurring estimate of £5.6 billion is below the £8.6 billion dividend before services, and the EU estimate of €21 billion is far below a €64.8 billion dividend. Borrowing against levy streams is not recurring funding. The first dividend is scheduled before the return assets could plausibly mature. Article 114 cannot safely be used to avoid fiscal unanimity, enhanced cooperation does not itself create an EU own resource, and Article 153 is not a QMV route to an EU compute dividend. - **Unsupported or invented premises:** data-centre capacity totals, country-by-country platform revenue attribution, €20 billion as available EIB equity capital, 4%-5% distributable returns, European Fiscal Board scoring power, platform take-rate authority and multiple Bill/mandate assumptions. - **Best repair:** retain only ordinary capital-income/buyback reform, market-valued public co-investment and national rent/supply policy; remove the EU dividend, capacity levy, borrowing and automatic platform caps. ## Convergence map Recurrence shows where independent proposals agree, not whether a mechanism is feasible. | Mechanism family | Approximate recurrence | Legal-fiscal reading | |---|---:|---| | Economic-agency dashboard and dormant triggers | 17/17 | Strong convergence. Best versions use several outcome indicators, cyclical adjustment and a separate fiscal certificate; weak versions trigger on AI adoption or self-reported attribution. | | Competition, portability, interoperability or exit rights | 17/17 | Strong and generally lawful only after designation/investigation or when directly linked to a public contract. It protects purchasing power but is not revenue. | | Rent-leakage defence through housing, energy or platform policy | 16/17 | Strong distributional insight. Supply, planning and energy measures remain national/local and must be costed; generic price caps are not automatically lawful or effective. | | Individual payment, capital account or transition account | 16/17 | Broad conceptual convergence. Fiscal closure, benefit interaction and residence rules separate credible designs from slogans. | | Public return from publicly supported technology investment | 15/17 | Strong if limited to separate finance/support agreements, market valuation, risk and State-aid review. Procurement is not a general equity-taking power. | | Shift from payroll dependence to property, capital, profit, rent or consumption bases | About 13/17 | Promising, but incidence and avoidance matter. Strong entries use a diversified mix; weak entries simply rename an AI tax. | | Compute-, AI-output- or AI-service-specific levy | 9/17 | Recurs often but is the least reliable family. Imported cloud use, heterogeneous hardware, attribution, double taxation and relocation are rarely solved. | | EU-level permanent dividend or tax route | 6/17 | Usually a failure mode. Direct tax and own resources require unanimity; the ordinary EU budget cannot run a deficit; Article 153 is not a shortcut. | ## Three strongest mechanisms across the set 1. **Outcome trigger plus fiscal-and-supply certificate.** Use several official measures of labour-financed agency for six to eight quarters, adjust for the cycle and measurement breaks, and permit a pre-costed payment step only after an independent institution certifies recurring coverage and delivery capacity. L, K and G provide the best variants. 2. **Two-ledger or three-ledger public-return discipline.** Public capital may earn public equity, convertibles, royalties or repayment rights when the state bears commercial risk. Capital, one-off proceeds, cost-reduction investment and recurring cash revenue must be kept separate. Only audited realised net returns after loss reserves may support a dividend. L, K, G, A and E converge here. 3. **The rent/absorption brake combined with lawful portability.** Test whether lower-income households retain the cash gain after housing and energy costs. Pause the next cash increase if scarcity captures it. In parallel, use contract-linked exit clauses and designation-based digital competition remedies to reduce lock-in. This is strongest in L, K and G. ## Three most common failure modes 1. **Fiscal illusion.** Entries count NWF capacity or EU mobilisation as revenue, assume high or immediate equity yields, omit eligible-population multiplication, double-count benefit savings, or promise services and cash from the same receipt. 2. **Unobservable or unstable bases.** AI-adjusted profit, AI share of output, autonomous outputs, accelerator-hours, foreign compute and platform “rent” are presented as if standard tax fields existed. Imported services, heterogeneous compute, outsourcing, nexus, treaty interaction, incidence and offshoring are usually unresolved. 3. **Institutional overreach.** Common errors include QMV EU taxation, Article 153 as a route to income or tax harmonisation, invented EU funds/registries, procurement conditions unrelated to the contract, automatic competition separation without due process and UK secondary legislation that cannot create the asserted power. A fourth recurring weakness is false entrenchment: a two-thirds threshold in an ordinary UK statute cannot bind a later sovereign Parliament. Transparency, property rights already accrued, independent audit and broad beneficiary coalitions are more realistic durability tools. ## Recommended narrow lawful integrated package The safest integrated core should be narrower than any union of the proposals. ### First stage: build the rails without a speculative AI tax 1. **Create a common economic-agency dashboard.** ONS, DSIT, Treasury and OBR in the UK, and Eurostat plus national authorities in the EU, should publish labour share, paid hours, median earned income relative to productivity, entry hiring and the wage-financed share of consumption. AI-use data remain contextual and retain their development caveats. 2. **Reform existing transition protection nationally.** The UK should replace the Universal Credit £16,000 cliff with a scored taper and a protected transition reserve, and pilot capped, cause-neutral earnings insurance for up to 18-24 months. Member states may make equivalent reforms. Training and adjustment funds remain temporary support, not permanent income. 3. **Use a strict public-return rule only where public capital or a scarce concession is actually supplied.** New grants, guarantees, concessional finance and co-investment may use equity, convertibles, royalties or repayable advances after market valuation and subsidy/State-aid review. Infrastructure concessions need express authority, transparent allocation and non-discrimination. Ordinary procurement remains confined to contract-linked evaluation, data rights, security, portability, continuity, exit and proportionate transition clauses. 4. **Separate three ledgers.** A capital ledger holds investments and losses; a supply/cost ledger funds housing, energy and digital-access investments; a recurring ledger contains only scored recurring tax receipts. NWF capacity and EU mobilisation never enter the recurring ledger. Realised public-investment returns can finance only an additional dividend after reserves. 5. **Use existing competition law lawfully.** The CMA may impose conduct or pro-competition remedies after Strategic Market Status designation and proper process. The Commission may enforce the DMA against designated gatekeepers. Neither regime permits automatic economy-wide API mandates or structural separation solely because a concentration ratio is crossed. 6. **Model a broad national tax migration.** The candidate mix is recurrent high-value property or land taxation, better alignment of substantial labour and capital income, taxable profits and distributions, and if necessary a destination-based consumption supplement with established import treatment. Publish incidence, decile effects, avoidance, migration and benefit interactions. Do not launch a compute, token, robot, AI-output or foreign-cloud levy. ### Second stage: a gated individual Agency Credit Use a two-key activation rule: - **Structural key:** at least three of five outcome indicators breach pre-set, cyclically adjusted thresholds for six to eight consecutive quarters. No individual must prove AI causation. - **Fiscal-and-supply key:** an independent fiscal institution certifies at least 110% recurring coverage over five years under stress assumptions; payment and appeal systems are tested; government publishes housing, energy and inflation effects. The first Agency Credit should be individual, taxable or progressively recovered from high incomes, disregarded for existing benefit withdrawal under a no-loss transition, and capped at a net envelope no greater than 0.5% of GDP. Later steps require fresh legislation or a pre-legislated schedule backed by named recurring revenue. If essential costs absorb a material share of the gain, the next cash step pauses and the marginal funding goes to supply, targeted services or competition enforcement. ### Division of legal responsibility - **United Kingdom:** tax, welfare, NWF governance, earnings insurance, property/housing choices and sectoral employment/due-process legislation. - **EU institutions:** common measurement, AI Act rights, DMA enforcement, procurement guidance, cross-border data/portability standards and State-aid control. - **Member states:** recurring transfers, social insurance, housing, energy and national tax mixes. - **EU funds:** EGF and ESF+ only within their existing training, employment, inclusion and adjustment mandates; never as an EU-wide permanent income source. ### Explicit exclusions The integrated package should exclude AI-output taxes, accelerator/FLOP levies, deemed foreign-compute charges, automatic structural separation, EU-level UBI under Article 153, speculative fund yields, borrowing against future levies for permanent benefits, procurement-based parent-company equity and statutory supermajorities presented as binding future Parliaments. ## Final judgment L, K and G provide the most legally and fiscally defensible architecture. M and N add useful rights and incremental sequencing. A and E contribute the best public-upside and anti-hollowing ideas, but only after their unsupported factual claims and missing income arithmetic are stripped away. The legally safest settlement is therefore not an “AI dividend” funded by an AI tax. It is a nationally financed, evidence-gated Agency Credit built on established broad tax bases, protected by a rent-absorption brake, supplemented by realised public investment returns, and supported by EU-level competition, rights, measurement, procurement guidance and State-aid discipline.