
Structural Reform of the French State
Public finances, institutions, implementation calendar and reform measures.
View the book →DELTA-SIERRA
Understand today, build tomorrow
Every amount should be reproducible. This page shows the starting baseline, the formula or verification method, and what still has to be deducted before a gross estimate can be called a net budget saving.
Rule: a transfer, redeployment, potential receipt or productivity gain is not booked as a saving unless expenditure genuinely disappears from the relevant budget or additional revenue is actually collected.
2026 National Assembly budget: 644.01 million euros; 577 members.
Mechanical upper-bound check: 2026 National Assembly budget × share of seats removed.
504.49 million euros is an upper bound that unrealistically assumes every budget line varies with the number of MPs. The announced 432–467 million euro range is lower, implying that some expenditure remains fixed. A net saving requires line-by-line separation of fixed costs, member-linked costs, staff, property, investment and transition costs.
Do not automatically count the upper end of the range before a full cost breakdown.
2026 Senate budget: 382.3 million euros; 348 senators.
Mechanical upper-bound check: 2026 Senate budget × share of seats removed.
The mechanical check lands inside the announced 232–250 million euro range. It still does not prove a net saving because permanent staff, property and residual commitments remain.
Keep as a parameterised scenario, not as an automatic saving.
In 2026, the legal ceiling allows up to 14 cabinet members for a full minister and 8 for a delegated minister (with a specific exception for Relations with Parliament).
A nine-ministry target mainly reduces political teams; it does not mechanically eliminate administrations that are reassigned to another ministry.
Separate cabinet savings, political functions, support functions and any reorganisation costs.
Model only after headcount/cost inventory; make sure the same saving is not counted twice with measure 1.08.
Absence-related deductions already exist: at the National Assembly, some committee absences can trigger a 25% deduction from the functional allowance; in the Senate, Rule 23 bis allows deductions up to the quarterly amount of the functional allowance.
The reform must therefore be assessed as an incremental effect against current rules, not against a fictional baseline with no sanctions.
The financial effect depends on future behaviour; the primary objective is attendance, not revenue generation.
Do not count 150 million euros as a deterministic saving; simulate only once the proposed rules are precisely defined.
National Assembly 2026: 83.80 million euros for parliamentary staff credits including payroll charges. Senate 2025: 68.283 million euros subsidy to AGAS, alongside ancillary resources.
Order-of-magnitude check: add the two available staff-credit envelopes and divide by two.
The check gives about 76.04 million euros, below the announced 85 million euros. Years, payroll charges, ancillary means and termination costs must be harmonised before the higher figure can be retained.
Build low/central scenarios after scope harmonisation; avoid double counting with 1.01 and 1.02.
DGCL 2025: 1,921 regional and territorial councillors on the broad perimeter, including single territorial authorities. 2026 scale: up to €2,877.37 gross/month for a regional councillor in a region of at least 3 million inhabitants.
Maximum check for base allowances only: number of councillors × maximum monthly allowance × 12 months.
This shows that 350 million euros cannot represent base allowances alone. Any larger institutional saving must separately document assembly overhead, support functions, premises, travel and other costs that genuinely disappear.
Do not count 350 million euros as one block; model allowances + assembly costs + support functions, with territorial double-counting check.
DGCL 2025: 4,041 departmental councillors. 2026 scale: maximum base allowance up to €2,877.37 gross/month depending on population.
A 50% reduction applied to the maximum base ceiling for every seat would give a theoretical bound of ≈69.8 million euros/year; this is not the net saving.
The 180 million euros target assumes a broader scope than allowances alone; assembly and support costs must be inventoried and overlaps removed.
Model by component, not with a single percentage.
Since October 2025, the legal ceiling is 14 cabinet members for a full minister and 8 for a delegated minister; the Plan target is 10.
For a full ministry already at the ceiling, the maximum gap is four cabinet posts; actual headcount must be measured cabinet by cabinet.
The measure becomes directly costable once cabinet headcount and remuneration are consolidated.
Make sure the same saving is not counted twice with the reduction in the number of ministries under 1.03.
The 2014 organic law already bars the combination of a parliamentary mandate with many local executive offices.
Any additional effect must therefore concern residual/horizontal combinations of local executive roles, with the functions precisely defined.
Define the operational perimeter before costing so that no saving is attributed to a prohibition already in force.
Consolidate the budget effect only after the residual perimeter is defined.
Local elected-official allowances are already governed by separate statutory scales according to mandate, function and population.
A “150% of the main mandate” ceiling requires representative multi-mandate cases and anonymised individual data to measure the real effect.
Build a microsimulation of combined allowances before including 80 million euros in a consolidated total.
Microsimulation; make sure the same saving is not counted twice with 1.09, 1.19 and seat-reduction measures.
DGCL 2025: 484,662 municipal councillors. A reduction of about 30% corresponds to roughly 145,399 seats on that count.
Most municipal councillors do not receive a standalone allowance; the financial effect cannot be obtained by multiplying removed seats by an average allowance.
Reconcile the 250 million euros estimate with municipal accounts: allowances actually paid, meeting costs, training, insurance, logistics and support functions.
Do not count automatically as a proven saving; build a model by municipality population band.
2026 CESE appropriations: 34.1 million euros.
The historical 36 million euros figure is close to the current envelope; 34.1 million euros is the 2026 gross ceiling before transition and residual commitments.
The live model will use the 2026 benchmark without rewriting the historical Plan figure.
Cap the scenario at the current envelope, then subtract transition costs.
Law caps a CESER member’s allowance at 45% of the maximum allowance of a regional councillor in the same region; the chair is capped at 50% of the maximum allowance of the regional president.
The legal rule allows allowances to be reconstructed, but a homogeneous national total for the 13 CESER bodies still has to be aggregated from regional budgets.
Break the 75 million euros estimate into allowances, staff, operating costs and shared costs.
Do not count before region-by-region aggregation.
Transparency measure: it first creates a disclosure obligation and a public information infrastructure.
No direct budget saving should be manufactured from a transparency effect.
Cost implementation separately; any audit/control savings should be measured ex post.
Structural effect, excluded from the savings total until a net effect is observed.
Gross monthly minimum wage from 1 June 2026: €1,867.02. Six times the minimum wage equals €11,202.12 gross/month (€134,425.44 gross/year), before defining the remuneration perimeter.
The saving requires the distribution of public remuneration above the threshold and a legal decision on whether bonuses, allowances and executive positions are included.
Define the assessment base and responsibility-related exceptions precisely before consolidating the 320 million euros figure.
HR microsimulation; make sure the same saving is not counted twice with 1.20.
A written parliamentary question published on 21 July 2026 specifically asks the Government for the 2023–2026 cost of resources and protection provided to former presidents.
The awaited 2026 breakdown should distinguish staff, premises, logistics and security protection.
The 5 million euros figure remains an estimate to consolidate until recent official data are published; security arrangements should remain threat-assessment based.
Do not count automatically as a proven saving; update as soon as an official answer is published.
2026 Presidency appropriation: 122.56 million euros.
The 30 million euros target is approximately 24.5% of the 2026 appropriation.
Separate staff, travel, security, property, investment and transferable functions before calculating a net saving.
Parameterised scenario; avoid unrecorded cost transfers.
Senate 2025 accounts: total restaurant subsidy €987,160, including €123,325 for the senators’ restaurant and €863,835 for staff restaurants. National Assembly catering services charge members, but a comparable net-cost perimeter still needs to be isolated.
The 12 million euros figure assumes a broader perimeter than the identified Senate subsidy; the actual net subsidised costs of both chambers and bars must be aggregated.
Publish net cost by establishment and beneficiary category before consolidating the target.
Do not count 12 million euros as one block; include only verified net subsidies.
“Benefits in kind” cover multiple schemes and public employers; some are function- or security-related.
An inventory must distinguish housing, vehicles, catering, travel, telecommunications, protection, former office-holders and schemes already counted elsewhere.
Build a scheme register before costing; justified security arrangements remain outside the removal scope.
Mandatory double-counting check against 1.16, 1.18 and mandate-related expenses.
RIFSEEP includes the IFSE and CIA components and coexists with specific allowance/bonus systems for certain civil-service corps.
The 450 million euros gain requires a map of senior civil-service bonuses, target populations and transition rules.
Define the base, target caps and transitional/acquired-right rules before consolidation.
HR microsimulation; make sure the same saving is not counted twice with the overall remuneration cap under 1.15.
Regions are expressly territorial authorities under Article 72 of the Constitution.
Count only expenditure genuinely removed, net of functions, staff, contracts and assets transferred elsewhere.
Phase in after decomposition; savings from councillors, CESER or regional services cannot be counted twice.
On 1 January 2026, DGCL counted 1,252 EPCI with own taxation: 21 metropolitan authorities, 14 urban communities, 230 agglomeration communities and 987 communities of communes.
Legal abolition of an EPCI does not abolish its functions. Separate governance/support genuinely avoided from transferred staff and services.
Model by EPCI type and function, including transfer costs.
The 128,000 staff figure is a transition pool defined by the Plan, not a standalone 2026 statistic.
Cost support, training, mobility and pay differentials; no saving is attributed to this measure itself.
Transition cost only; savings belong to posts later genuinely removed or left unfilled.
In the Plan this line is classified as a productivity gain, not a direct budget saving.
Document administrative time saved, compliance costs and delays. Do not automatically convert 8–12 billion euros into expenditure cuts.
Separate productivity module until appropriations or jobs actually fall.
A one-in, two-out rule acts on regulatory flow; by itself it creates no immediate saving.
Publish rules created/removed and measured burden; convert to euros only where separately observed.
Outside direct savings total.
Légifrance maintains a live list of codes. The figure 69 is the Plan's historical reference and must be re-frozen at a precise date before use as a current baseline.
Measure simplification, duplication removed and recodification cost; fewer codes do not automatically mean lower spending.
Structural effect, with legal transition cost separated.
This line is classified as productivity gain in the Plan. Digitalisation also requires systems, cybersecurity, maintenance, support and training.
Separate the 10–15 billion euros potential from observed fiscal savings; any workforce reduction passes through 2.09.
Productivity plus digital costs; no automatic addition to payroll savings.
France services already centralises procedures and handles about 1.2 million assisted interactions per month.
The one-stop shop should extend/integrate an existing base; assess coverage, connected services, staff, premises and genuinely duplicated counters.
Savings only for counters genuinely removed while accessibility is maintained.
In 2024, 136,700 new direct pensions were awarded to people who had worked in public service. A mechanical 30–50% screen equals 41,010–68,350 posts; retirements and all natural departures are not identical.
Calculate post by post: observed departure, replacement need, priority service, employer cost and vacancy date. At end-2024 public service had 5.8767 million staff (+0.6% year on year).
Only a genuinely unfilled vacancy produces a saving; mobility, severance or automation cannot add a second saving on the same post.
Supported mobility moves a resource to a priority service; by itself it does not reduce payroll.
Track origin, destination, avoided vacancy and priority need; the effect may be an avoided future hire.
No direct saving; any avoided hire is counted once in the workforce scenario.
The decree of 6 August 2026 made mutually agreed severance permanent. Current minimums range from one sixth to one third of a month's gross pay per year depending on seniority; one month per year is therefore much more generous.
Model the upfront cost by seniority and pay, then compare with employer cost avoided only if the post genuinely disappears.
Upfront investment; any later saving passes through 2.09, never twice.
Progressive retirement for public employees is already available from age 60 subject to conditions. The age-58-to-62 proposal must isolate what is new for ages 58–59 from what overlaps existing rules from age 60.
Cost work percentage, pension fraction, contributions, any replacement and knowledge transfer.
Separate transition cost; payroll reduction only net of replacement and free of double counting against 2.09.
DGAFP has already launched AI training and opened cross-branch framework negotiations in June 2026. The 400,000-person target is therefore a scale-up to be costed.
Cost training, staff time, certification, tools and support; gains appear only after measuring automated tasks and resources actually redeployed/not replaced.
Upfront investment, no automatic fiscal saving; productivity effects connect to 2.07 and category 09.
The Senate's 2025 perimeter covers 1,153 national public bodies, including 426 budget operators. Operators receive 77.0 billion euros in public funding, but this includes grants, transfers and earmarked taxes and therefore is not an overhead figure.
Keep the Plan's 552 retained entities as a working perimeter, then audit payroll, property, procurement, communications, IT, governance and intervention spending separately. Never apply a savings rate to all funds administered.
No automatic aggregate saving. Only itemised, free of double counting overhead reductions may enter the model.
The Senate report qualifies the idea of a general agency pay premium: the average monthly gross pay of the operator executives' 'top 10' was €7,986.58 in 2023 versus €16,799 in ministries, although a few operators exceed their relevant ministry benchmarks.
Identify operator by operator the remuneration above the chosen cap and calculate savings only on pay actually reduced to that cap. Do not apply a blanket cut to all executives.
Direct saving only for verified individual excess above the cap; make sure the same saving is not counted twice with measures 1.15 and 1.20.
The Senate estimates operator support functions at about 2.7 billion euros in 2024 on its perimeter excluding universities and research centres. A 20% cut, described as highly ambitious, gives a reference screen of about 540 million euros.
Use 540 million euros as a parliamentary control benchmark, then rebuild savings operator by operator. Remove gains already captured by merger, abolition, digitisation or non-replacement.
May enter the budget only after entity-level allocation and overlap removal.
The Senate report records 402,489 FTEs under operator employment ceilings and 77,058 outside them. This is a transformation perimeter, not an automatic AI saving.
Measure time released task by task. Convert it into a budget saving only when procurement, recruitment or a post is actually avoided, reduced or left unfilled.
Separate productivity module until fiscal realisation; mandatory double-counting check against 2.07, 2.09, 2.13 and the AI package.
The Senate notes that operator mergers and abolitions from 2015 to 2019 produced only a 0.8% fall in jobs at constant scope. A legal merger therefore does not translate into proportional savings.
Review all 35 clusters individually: governance removed, pooled support, property, IT, pay-harmonisation and transition costs. The 2.3–4.1 billion euros range remains a historical target to consolidate cluster by cluster.
Enter only each merger's net gain after transition costs and double-counting check against 3.03/3.04/2.09.
The Senate report recommends several reinternalisations but stresses that the missions continue. Reintegration changes the administrative vehicle; it does not mechanically remove intervention appropriations.
Value only legal-entity, governance, duplicate oversight, support and property costs avoided, net of costs transferred back to ministries. The historical Category C potential must be rebuilt entity by entity.
Net restructuring only; transferred mission funding remains public expenditure unless a separate policy decision changes it.
ADEME's 2024 main budget recorded 1,029.5 million euros in revenue, of which only 60.6 million euros was own-source revenue. The Senate stresses that retained aid schemes would still cost the public budget and already recommends direct State-to-region transfers for some funds.
Strictly separate retained aid, legacy commitments, transferred expertise and overhead genuinely removed. ADEME's budget is not treated as a saving.
Only net overhead removed, after transfer costs and legacy commitments.
OFB reports a 623 million euros initial 2024 budget, funded 70–80% by the water agencies; 73.5 million euros is contributed to the eleven national parks. These flows therefore cannot be equated with OFB overhead.
Separate environmental enforcement, expertise, biodiversity, parks, grants and support functions. Any transferred mission keeps its cost in the receiving body or service.
Net overhead saving only; transfers to parks and retained missions excluded from savings.
For 2026, AFIT France earmarked revenue and appropriations are estimated at about 3.75 billion euros. Yet the Senate describes a structure with four permanent staff relying entirely on central government and recommends reinternalisation.
The 3.75 billion euros funds infrastructure and does not disappear with the legal entity. Value only governance, accounting, own overhead and budget simplification net of costs taken back by the State.
Reinternalisation with limited overhead savings; infrastructure appropriations are never treated as savings.
Budget funding for regional health agencies under programme 155 is set at 627 million euros for 2026. The Senate report recommends transferring their responsibilities to decentralised State services at regional and departmental levels.
Transferring missions means staff, property, systems and health functions must be taken over. Costing must focus on duplicate and governance costs removed, not the entire envelope.
Net gain after transfer costs, free of double counting against 2.01, 2.03, 2.09 and health reforms.
DRAAFs are decentralised State services rather than autonomous aid envelopes. Their reform directly overlaps abolition of the regional administrative tier and staff mobility.
Keep the historical 60–120 million euros range as a prudent estimate to consolidate through payroll, property, transferred functions and the actual number of posts avoided.
No model entry before double-counting check against 2.01/2.03/2.09/2.10 and territorial allocation.
The Senate identifies the water agencies among major intervention operators, with about 1.9 billion euros in intervention charges. It also notes that DDT/DREAL services perform operational tasks for other operators. These funds are not homogeneous overhead.
Split into three sub-perimeters: DDT services, DREAL services and water agencies; isolate missions, fees, grants, enforcement/control and support. Make sure the same saving is not counted twice with OFB and regional-tier reform.
Enter only by sub-perimeter, net of transferred missions and flows.
The 2025 Senate report counts 317 consultative bodies on its perimeter. It estimates their real total cost closer to 50 million euros than the 30.9 million euros documented for 2023, and puts a 20% reduction at about 10 million euros.
The historical 316-body perimeter and 120–240 million euros target must be reconciled with the Senate definition before consolidation. Build a committee-by-committee register and aggregate remuneration, secretariat, travel and staff time actually avoided.
Enter only documented annual costs of bodies actually abolished; staff-time gains remain separate until expenditure falls.
The CSA has not existed since 1 January 2022: it merged with Hadopi to form Arcom. The 2026 draft budget provides 50.52 million euros for the Arcom action and an employment ceiling of 378 FTEs.
Keep the historical wording in the Plan table, but audit the measure against today's Arcom. The 85 million euros target cannot simply be mapped onto a 2026 50.52 million euros envelope; missions to abolish, transfer or pool must be specified.
No saving above net costs genuinely removed; transferred-mission costs deducted.
The 2026 Finance Act opens 2,425.577 million euros for France Télévisions. This envelope is a reference bound: privatisation would also create transition costs and potentially one-off sale proceeds.
Separate the annual grant reduction, any retained public-service obligations, social/contractual commitments and sale proceeds. Sale proceeds are one-off and must never be added to annual savings.
Recurring saving = net reduction in public funding; disposals go to a separate one-off account.
The 2026 Finance Act opens 648.033908 million euros for Radio France. A mechanical 30–45% reduction equals roughly 194.4–291.6 million euros, close to the Plan's historical range, without prejudging operational feasibility.
Turn the percentage into an operational programme covering stations, transmission, property, production, digital, staff, retained missions and timing. Costing must be consolidated against an explicit public-service scenario.
Net grant reduction only after transition costs; make sure the same saving is not counted twice with any wider public-media reform.
For 2026, direct budget support to the written press is set at 178.29 million euros. The Senate adds 65 million euros in tax expenditure, including 58 million euros for the super-reduced VAT rate. These two components have different fiscal natures.
Replace the 320 million euros shortcut with a full inventory of direct aid, tax expenditure and any other schemes. Removing tax expenditure is a potential revenue increase, not a spending cut, and must be modelled separately.
178.29 million euros theoretical maximum on direct aid before exceptions; 65 million euros in the tax-expenditure module, free of double counting against 6.11.
The Senate uses a perimeter of 1,153 national public bodies, excluding basic social-security schemes and other bodies governed by the Social Security Code. This count must be dated and versioned rather than treated as permanent.
Create a versioned public register: identifier, legal status, supervising ministry, mission, staffing, funding, creation, merger, abolition and last verification date. Every perimeter change must be traceable.
No direct budget gain; this is control infrastructure for scope and double counting.
The 2026 Finance Act opens 12.967 billion euros in payment appropriations for the Justice mission, including 4.742 billion euros for judicial justice and 5.505 billion euros for prison administration.
The word “doubling” must be tied to a stable accounting perimeter. If it applies to the entire 2026 Justice mission, steady-state expenditure would rise by nearly 12.97 billion euros per year: this is public investment, never a saving.
Additional programmed expenditure by year; productivity effects and any revenue remain in separate modules.
The reference network comprises 164 judicial courts. The EU AI Act treats systems used by a judicial authority to help research or interpret facts and law as high-risk when they materially influence decision-making; final decision-making must remain human-driven.
Treat 4.02 as the umbrella for uses 4.03 to 4.05. The 1.2 billion euros figure must not be added in full to the gains attributed to its sub-uses. Deployment should be staged with bias testing, security, traceability and mandatory human control.
Productivity module only until realised through shorter delays, avoided recruitment or capacity actually released.
The CEPEJ identifies case routing and management among possible uses, while requiring quality, security, transparency and user control. A narrow procedural task has a different risk profile from a tool influencing legal analysis.
Measure clerk and judge time released by case type; no automated adverse decision. The gain is free of double counting against 4.02 and general public-administration digitisation.
Productivity; fiscal conversion only when staffing, temporary work, procurement or recruitment is actually avoided.
The CEPEJ explicitly identifies legal research as a possible assistive use. The AI Act specifically addresses systems helping judicial authorities research and interpret facts and law.
Measure research-time reduction without weakening reasoning duties or source verification. Citations proposed by the tool must remain verifiable and versioned.
Productivity included in 4.02 unless a distinct perimeter is demonstrated.
The AI Act distinguishes systems detecting patterns or deviations in already-completed decisions when they neither replace nor influence the human assessment without proper review. CEPEJ principles require transparency, non-discrimination and user control.
Retrospective quality-control and research use: documented samples, non-prescriptive indicators, no automatic individual scoring of judges and no automatic reopening of a decision.
No automatic budget gain; quality and risk indicator only.
For direct taxes, turnover taxes and similar levies, litigation following an administrative claim is mainly handled by administrative courts. Costing must therefore separate administrative justice, any judicial tax litigation, and prior processing by the tax administration.
Define precisely which simple disputes can be resolved administratively, through mediation or digital procedure without removing access to a court. The 2–4 billion euros range remains a productivity potential to consolidate using case volumes and cost per case.
Separate productivity module; no saving until volumes, competent jurisdiction and full cost per case are established.
Budget documents cited by the Senate report an average civil-processing time falling from 8.2 months in 2021 to 7.0 months in 2024; 84.8% of first-instance civil cases ended within twelve months in 2024. Some case types nevertheless remain at 14.9–18.9 months.
The six-month target becomes a public-service performance objective measured by case type, without artificial acceleration through case rejection or weakening adversarial rights.
Outcome indicator; it becomes a euro value only through measures that actually produce the improvement.
The 2023–2027 Justice Programming Act already provides for 1,500 additional judges, 1,800 court clerks and 1,100 justice attachés by 2027. For 2026, the Judicial Justice programme plans +660 FTEs, including 286 judges and 342 court clerks.
Separate recruitment already funded under the existing trajectory from the Plan’s 5,000-post target. Redeployment is not a saving: it moves payroll. Every created post must be funded or offset by a saving actually realised elsewhere.
Payroll cost or redeployment; budget-neutral only if the source post genuinely disappears.
The EU AI Act states that AI tools may support judges’ decision-making power but should not replace it: final decision-making must remain human-driven. The CEPEJ likewise applies a user-control principle.
Make this a cross-cutting safeguard: identity of the validating officer or judge, date, system version, material consulted and explicit ability to disagree with the algorithmic suggestion.
Non-monetised safeguard; IT and organisational cost included in AI deployment cost.
European judicial-AI frameworks emphasise transparency, quality, security, auditability and human control. The AI Act also imposes specific requirements on high-risk systems, including risk management and traceability.
For each assisted process archive model version, supplied sources or documents, relevant outputs, validator identity and access logs, with security, confidentiality and retention rules to be expert-reviewed under archives and data-protection law.
Compliance and infrastructure cost; no direct saving.
DREES counts about 1.4 million hospital employees at end-2024. Administrative staff represent 11% of the workforce, roughly 155,000 people as a mechanical order of magnitude. This category also includes medical secretaries, reception, dispatch and billing staff.
The measure simultaneously calls for less administration and more care staff: payroll redeployed to care cannot also be counted as a budget saving. The 3.5 billion euros figure therefore has to be rebuilt facility by facility after protecting essential functions and transition costs.
Net saving = administrative costs actually removed − care-staff reinvestment − transition − transferred functions.
HAS distinguishes medical-purpose AI from support-function tools that may reduce administrative workload, including assistance with medical summaries. It calls for supervised, deliberate and controlled use.
Voice reporting can return time to care, but an hour released is not a euro saved. The 2.8 billion euros figure remains productivity potential until an actual cost is avoided. It must be free of double counting against 5.01 and the cross-government AI strategy.
Time released by occupation and unit; fiscal conversion only after a measurable organisational change.
French National Health Insurance publishes extra-fee amounts from the SNDS by profession and territory. In 2024, 51.7% of eligible sector-2 specialists belonged to Optam/Optam-ACO; the rate reached 53.1% at 30 June 2025.
Extra fees are private professional income, not existing public revenue. Any recovery mechanism requires a legal basis, tax base, threshold, rate, collection procedure and behavioural analysis. The 1.2 billion euros figure must be recalculated from national SNDS data before inclusion.
Gross revenue − behavioural effects − collection costs; no double counting with 5.08 on the same fee base.
Waiting times primarily reflect actual clinical capacity. A rule targeting platforms must distinguish lack of available slots, misleading information, waiting-list management and the responsibility of the practitioner or local care supply.
The measure creates no direct budget saving. Operational drafting must avoid reducing patient information through delisting; priority should be transparent waits, waiting lists and interoperability.
No automatic fiscal effect; access indicators only.
The 2026 budget-preparation benchmark showed 8,114 FTEs for regional health agencies and 627.142 million euros in operating subsidy. Separately, the 2026 Social Security Financing Act sets the regional intervention fund and national investment sub-target at 6.4 billion euros, while the 2 June 2026 order sets the health-insurance contribution to the regional intervention fund alone at 5.178 billion euros.
Billions in regional intervention funding finance territorial health policy and are not the administrative cost of the agencies. The 180–350 million euros target must be tested only against removable governance, support and legal-autonomy costs, less the capacity required by the ministry or prefectural structure taking over the missions.
Structural saving only; transferred health-policy funding remains outside savings unless a policy is explicitly discontinued.
Digital health-data hosts must hold HDS certification. The French Digital Health Agency counted 391 certified HDS hosts in May 2026 and 9 authorised certification bodies.
This is not a saving: it is a security floor. Any health AI or cloud architecture must preserve HDS certification, data protection, traceability and the location requirements of the applicable framework.
Compliance and cybersecurity cost, never an automatic saving.
HAS already recognises digital medical devices and AI systems used to assist screening, diagnosis or clinical decision-making. Evaluation remains based on clinical benefit, safety and appropriate use.
No saving is assigned by default. Each use must pass the applicable clinical and regulatory evaluation, with professional validation, false-positive/false-negative monitoring and organisational impact measurement.
Separate clinical scenario: technology costs, tests avoided or induced, outcomes and quality.
The 2024-2029 medical convention already uses Optam/Optam-ACO to moderate extra fees. In 2024, 51.7% of eligible sector-2 specialists participated, rising to 53.1% at 30 June 2025.
The cap is primarily an affordability measure. Its fiscal effect depends on the cap level, reimbursement rules and behaviour. It must be modelled jointly with 5.03 on a common fee base to avoid double counting.
Separate patient / statutory insurance / complementary-insurance effects; no automatic public saving.
INSEE measures compulsory levies at 43.6% of GDP in 2025. A 50% constitutional ceiling would therefore not bind today; it would act as a safeguard against future increases.
Do not confuse the 50% legal ceiling with the separate long-run 30% strategic target. The first limits how high levies may rise; the second requires actual savings and a financing lock.
€0 automatic effect. Each tax cut is simulated separately and blocked if it recreates a structural deficit.
Since 1 January 2026, the single degressive general reduction is already maximal at the minimum wage and tapers off up to below 3 minimum wages. Any further reduction therefore starts from an already broad relief system.
A contribution cut first reduces social-security revenue. Employment, net-wage and employer-cost effects must be microsimulated before any tax feedback is counted. No debt financing.
Gross revenue cost − observable fiscal feedback + employment effect; kept separate from gains to employers and workers.
2026 law: €100,000 allowance per child and €15,932 between siblings. decision: 1 million euros per child, €500k for a qualifying blended-family child with the reference tax burden divided by three above the threshold, and 1 million euros per sibling.
Microsimulation of net annual revenue cost by relationship and estate band; legal definition of blended families; interaction with prior gifts and effective date. The measure remains a revenue loss that must be funded; it is not a saving.
Microsimulate three categories: direct child; qualifying blended-family child; sibling. Compare 2026 law with .
The current regime fully exempts income tax after more than 22 years of responsible measureship and social levies after more than 30 years. Ten years would therefore accelerate exemption sharply.
The cost must be simulated using transactions, holding periods, gains, already exempt principal residences and any market-unlocking effect. The later fifteen-year strategic scenario is shown separately from the ten-year source measure.
Lost income-tax and social-levy revenue, adjusted for transaction-volume effects.
In 2026, the tax is declared within 90 days after completion; above €1,500 it can be paid in two instalments, at 90 days and nine months. Spreading it over fifteen years fundamentally changes local-government cash flow.
This is not a saving: it is implicit credit to the taxpayer, with financing, inflation, default and administration costs. For new housing, 6.05 and 6.13 cannot both apply to the same tax.
Present value of local receipts + administration + default; mutually exclusive with 6.13 for new housing.
Budget documentation estimates the total cost of the research tax credit at 8.041 billion euros in 2026. This is an absolute ceiling, but it does not by itself validate the 5.5 billion euros attributed to large companies.
Before consolidation, a DGFiP breakdown is needed by company size, tax group, R&D spending band and credit actually used. Amounts redirected under 6.08 cannot simultaneously count as deficit reduction.
Gross revenue recovered − R&D behavioural effect − any 6.08 redirection. No double counting with 6.11.
The total research tax credit is estimated at 8.041 billion euros in 2026 and the innovation credit at 230 million euros. A 2 billion euros yield depends entirely on the rule chosen for SMEs and mid-caps and cannot be inferred from the total cost.
The website will require a parameterised scenario: rate, base, ceiling, group treatment, SME/mid-cap status and eligible spending. Depending on design, the measure may raise revenue, cost revenue or simply redistribute support.
Reform scenarios compared with current law; no automatic 2 billion euros.
The Research and Higher Education mission represents more than 31 billion euros in 2026 payment appropriations according to budget documentation. Redirecting recovered research-tax-credit revenue is new public-research spending.
Each euro allocated to 6.08 reduces by the same amount the net deficit benefit from 6.06/6.07, even if it improves scientific and economic potential.
Explicit investment spending linked to revenue actually recovered.
In 2025, tax audits assessed 17.1 billion euros in duties and penalties and DGFiP collected 11.4 billion euros. This is the starting point: AI cannot claim as a gain amounts already detected or collected without it.
The model measures only the increment attributable to AI, net of false positives, litigation, IT and staffing costs. Assessed, collected and prevented amounts remain separate.
Causal additional cash collection − AI cost − audit cost − litigation. Free of double counting against 9.06.
In 2025, nearly 3.1 billion euros of social fraud was detected, prevented or reassessed. These concepts are heterogeneous and do not all equal cash collected. For example, Urssaf reassessed 1.503 billion euros for undeclared work while collecting about 100 million euros during the year.
The AI gain must be incremental to existing controls and split between prevented benefits, recovered contributions, overpayments, penalties and actual cash collection.
Prevented loss + additional cash collection − programme cost; no gross addition with 11.05.
Budget documentation lists tax expenditures measure by measure. The 2026 budget bill already contained several removals and rationalisations; Delta-Sierra must therefore start from the 2026 inventory rather than an arbitrary aggregate percentage.
Each tax expenditure must have: budget cost, beneficiaries, objective, evaluation, behavioural effect, EU-law constraints and exit scenario. The research tax credit is removed from 6.11 when treated separately under 6.06/6.07.
Measure-by-measure sum after removing items already counted elsewhere and behavioural effects.
The 30% figure appears in some procurement rules for matters unrelated to supplier profit margins, including advances or defence-contract subcontracting. It is not a general supplier-margin cap today.
Measure 6.12 directly overlaps 8.04 and must become one consolidated line. To verify 3–5 billion euros, the margin definition must be specified, an open-book sample built, risk and competition considered, and actual purchase-price reductions measured.
Saving = counterfactual price − post-reform price − control costs; 6.12/8.04 counted once.
Development tax is a local tax used in particular to finance public infrastructure required by development. Removing it on new housing lowers project cost but reduces local-government revenue unless compensated.
The website separates buyer/developer gain, lost local revenue, any state compensation and construction response. For housing exempt under 6.13, the 6.05 payment-spreading measure no longer applies.
Lost local revenue − any additional activity/revenue; mutually exclusive with 6.05 on new housing.
ARENH ended on 31 December 2025 and the VNU applies from 2026. CRE estimated EDF’s 2026 nuclear revenue at 23.7 billion euros (€65.86/MWh), while the 2026 VNU rebate was set at €0/MWh.
The source 3–5 billion euros scenario is preserved, but it can reduce the public deficit only through an explicit channel: dividends, tax, transfer or lower support needs. EDF must also finance maintenance, fuel, grid and investment.
Separate EDF account: nuclear-revenue change → costs → profit/cash flow → possible dividends/taxes. Consumer account kept separate.
France’s post-ARENH framework now rests on the VNU, with thresholds and a rebate defined by the applicable statutory and regulatory framework.
No amount is added until renegotiation is translated into concrete parameters: thresholds, base, beneficiaries, interaction with the European market, and effects on EDF and consumers.
€0 by default; impacts captured through 7.01 once precise legal parameters are defined.
The ministry states that French oil production represents about 1% of consumption and recalls the progressive phase-out framework stemming from the 2017 law.
Repealing a ban creates no revenue by itself. Revenue arises only after permits, a commercially exploitable discovery, investment, production and taxes/royalties actually collected.
€0 in the baseline; revenue only from documented production, net of public costs.
Current domestic production remains marginal relative to consumption. An exploration programme therefore begins with geological, technical, administrative and environmental costs.
Separate exploration cost, private/public financing, discovery probability, development lead time, royalties and taxation. No hypothetical field enters the fiscal balance.
Negative short-term investment if publicly funded; later revenue only after production begins.
CEA states that ASTRID ran from 2010 to 2019 and ended after the demonstrator was deferred. Work on fast reactors and the programme’s knowledge base continues in other forms.
A restart is first a multi-year R&D, engineering, safety and capital programme. No future industrial benefit is counted as fiscal revenue before a documented business case exists.
Explicit investment cost; €0 automatic revenue/saving over five years.
In May 2026, NASA awarded $6.9m to Interlune to develop and validate lunar-resource prospecting technologies including helium-3. This confirms a real R&D topic, not near-term exploitable revenue.
Treat as research/strategic option. No helium-3 revenue enters the five-year budget until prospecting, extraction, transport, industrial demand and fusion economics are established.
€0 positive fiscal effect; any R&D cost isolated.
The ministry petroleum-market document reports an annual-moving road-fuel volume of 47.428 billion litres around August 2025. A uniform €0.20/l cut on every litre gives a gross mechanical screen of about 9.49 billion euros per year. Road-fuel demand was still down 3.4% year on year in Q1 2026.
The Plan’s historic 6 billion euros cost remains the source value. A current estimate must specify eligible litres, professional refunds, excise/VAT interaction and behavioural response. No cut is debt financed.
Gross revenue loss + VAT/volume effect − any economic feedback; funding required before activation.
The seven main concessions, covering more than 90% of the concession network, expire between 2031 and 2036. In 2024, concession companies generated 12.8 billion euros revenue and 4.3 billion euros net profit. ART also identifies end-of-contract obligations: about 1.2 billion euros extra maintenance and 0.4–5.1 billion euros of investments to be clarified.
No 7.08 revenue before contractual expiry in the baseline. After handback: toll/other revenue − operations − maintenance − investment − financing. Current private net profit is only a sector-scale bound, not guaranteed public revenue.
Activation from 2031 at the earliest; net post-concession scenario. Free of double counting against 7.09 and 7.10.
A long-distance transport-infrastructure operating tax already exists: it applies above 120 million euros of revenue when average profitability exceeds 10%, at 4.6% of revenue above the threshold.
The Plan’s 1.2 billion euros cannot simply be added to current receipts without defining the new base and rate and subtracting existing tax revenue. As concessions expire, the private taxable base shrinks or disappears depending on 7.08/7.10.
Incremental revenue only: reformed tax − existing tax; gradual phase-out with the relevant concessions.
ART explicitly calls for decisions before 2031 on future management arrangements: operating model, perimeter, investment, financing and any future contract duration.
7.10 does not create an independent amount: it is the institutional architecture for 7.08. The “Italian model” wording is preserved as the source label; any operational comparison must identify the exact operator, legal framework and comparable performance.
€0 standalone; costs and revenue included in the 7.08 post-concession scenario.
The Consumer Code already regulates the “fait maison” label: a dish must be prepared on site from raw products, subject to regulatory exceptions. In June 2026, the National Consumer Council adopted an opinion preparing a reform of the scheme.
The measure should be assessed through compliance rates, consumer clarity and enforcement costs. No amount enters the savings counter without an identifiable expenditure reduction.
€0 default budget effect; any enforcement cost is modelled separately.
EGAlim already requires at least 50% sustainable and quality products, including 20% organic. In 2024 reported purchases, the observed rates were 29.5% sustainable products and 11.8% organic; 33,797 sites were covered, about 40% of all sites. “France” origin can now be reported voluntarily.
A quota directly based on French origin requires legal redesign: EU public procurement requires equal treatment and non-discrimination and generally prohibits favouring a specific origin. The scenario must therefore test lawful quality, sustainability, seasonality, environmental-performance or direct-supply criteria.
No automatic saving; model purchase-price differential, logistics, waste and agricultural effects.
The proposal potentially spans planning rules, direct sales, home processing, small-scale livestock, gardens and health rules. No single budget aggregate maps cleanly to this heterogeneous scope.
Break the measure into legal sub-reforms before costing it. Potential gains first concern household income, resilience and local activity; revenue losses or enforcement costs must be separated.
€0 by default; sector modules only after precise definition.
Measure 6.12 already contains the same mechanism (“cap public-procurement margins at 30%”) and was audited in . The 3–5 billion euros amount must therefore be tested only once against procurement data.
Measure to which the saving is assigned: 6.12. Line 8.04 remains visible to preserve the source corpus but contributes €0 additional amount to any consolidated total.
Alias of 6.12; the same saving cannot be activated twice.
Development tax is already covered by measures 6.05 (payment spreading) and 6.13 (abolition for new housing). 8.05 extends the issue to business creation but uses the same local tax base.
Any reduction must first be costed as a revenue loss for local authorities before estimating any additional investment or activity. The same taxable floor area cannot be counted simultaneously under 6.05, 6.13 and 8.05.
Separate local-tax cost; mutually exclusive scenarios by construction type.
Since 1 January 2023, all business formalities already go through a single online portal operated by INPI, replacing six networks of business-formality centres. Delta-Sierra must therefore measure what it simplifies beyond this existing baseline.
Measure forms removed, processing times, required documents, human interventions, user cost and administrative cost. Time saved affects the deficit only after actual budget or staffing reductions.
Separate productivity account; budget conversion only through non-replacement, lower outsourced services or appropriations.
The Senate estimates aid to all businesses at at least 211 billion euros under a broad definition and 108 billion euros under a strict definition in 2023, but states that the amount received specifically by large companies cannot currently be known precisely. Corporate-tax credits represented 7 billion euros in 2023, of which 4.3 billion euros was concentrated on large companies, but this component already overlaps the Plan’s tax measures.
Never apply an arbitrary percentage to the 211 billion euros figure. Build a company-by-company and scheme-by-scheme register, distinguish grants, tax expenditures, social-contribution relief, loans/guarantees and EU aid, then remove schemes retained or already counted elsewhere.
€0 by default until a consolidated large-company register is built; activate line by line using the net amount actually abolished.
The 2025 Senate report notably recommends full repayment of State or local-authority aid when the site or activity that justified it is relocated within the following two years, and ex-ante definition of other repayment clauses.
Conditionality does not create a saving on enactment. It may reduce future aid or generate clawbacks only when conditions are breached. Aid already abolished under 8.07 cannot be counted again.
€0 by default; net recovery only for documented and collected events.
Precedents already exist: students at the Écoles normales supérieures are subject to a ten-year professional commitment with possible reimbursement upon breach; École Polytechnique also has a reimbursement regime and, depending on the category of graduate, service obligations that can reach ten years.
Extending the rule to all “grandes écoles” requires defining covered institutions and students, the public consideration received, eligible activities and compatibility with professional and EU freedoms. Training cost cannot be presumed fully recoverable.
€0 revenue; retention and service indicators only.
ENS and Polytechnique schemes show that repayment can be organised where a prior commitment exists and its calculation, exemptions and procedure are defined. These schemes already generate claims within their own scope and cannot be counted again as “new” revenue.
Expatriation alone does not create a claim. The triggering event, recoverable public cost, duration, prorating, exemptions and cross-border collection must be defined. Consolidated revenue is the additional net cash collected beyond existing schemes.
€0 by default; net cash only after a valid claim and actual collection.
The State is no longer starting from zero: DINUM already operates an interministerial AI stack, including Albert API, while the Cloud-at-the-Centre doctrine governs hosting. Albert API mutualises several models and is already used in more than 70 public projects, with more than 100,000 weekly requests.
9.01 becomes the umbrella architecture. Savings or revenue are assigned to the use cases that actually materialise them; the architecture itself cannot count DGFiP, justice, health, agency or civil-service gains a second time.
€0 direct; investment and operating costs are carried by 9.03, 9.08 or the owning business projects.
The Cloud-at-the-Centre doctrine already requires particularly sensitive data hosted in commercial cloud to use SecNumCloud or at least equivalent European qualification, with protection against extraterritorial access. ANSSI maintains the SecNumCloud 3.2 baseline and a catalogue of qualified or qualifying offers.
The measure is broader than today’s baseline if it requires SecNumCloud for every use. Sensitive/strategic data, admissible internal clouds, European equivalents, capacity and price differentials must be defined. No automatic saving.
€0 saving; migration cost + hosting OPEX delta, without double counting HDS or business platforms.
Albert API already constitutes an interministerial multi-model inference platform: unified API, open-weight or partner models, RAG, OCR, classification and sovereign hosting. The source measure must therefore be audited as scaling/industrialising an existing baseline, not creating it from scratch.
Cost accelerators/GPUs, hosting, possible licences, operations, cybersecurity, support and scaling. Gains from applications using the platform remain assigned to their business lines.
Investment/OPEX to be consolidated; €0 own saving.
The Cloud-at-the-Centre doctrine already encourages diversity of technologies, suppliers and infrastructures for continuity and recovery. Albert API uses a standardised interface compatible with OpenAI conventions, which improves technical portability without eliminating exit costs.
The 12-month period should become an exit-plan requirement: export of data, models/configuration, documentation, switchover tests, supplier assistance and capped exit fees. The benefit is primarily avoided risk.
€0 by default; avoided exit costs only after contract benchmarking.
The population baseline is now consistent with official statistics: 5.8509 million employees excluding subsidised contracts at end-2024, including 2.5840 million in the State civil service. The source cost of 3.2 billion euros/year for five years is 16 billion euros total, about €2,735 per employee over five years (about €547/employee/year). A three-branch civil-service AI framework negotiation started in June 2026.
Split teaching cost, paid training time, certification, replacement, infrastructure and support. Measure 2.13 (400,000 staff) is a potential subset: its cost must be subtracted if both lines are activated.
Source scenario: −3.2 billion euros/year for 5 years, to be replaced by bottom-up costing; mandatory double-counting check with 2.13.
DGFiP has about 95,000 staff. In 2025 tax audits notified 17.1 billion euros, collected 11.4 billion euros, and data mining/AI enabled 2.8 billion euros to be recovered. The 15 billion euros lower bound therefore already exceeds total annual cash collections from 2025 tax audits: the source figure must be treated as long-term transformation potential, not immediately available incremental revenue.
Measure to which the saving is assigned: 6.09. New revenue is additional AI-enabled cash collection above the existing baseline, net of costs, litigation, displacement and false positives. 9.06 describes the deployment domain and does not add another 15–25 billion euros to the total.
€0 additional: revenue alias of 6.09; DGFiP productivity is separate if it actually reduces costs.
521,000 positions represent about 8.9% of the 5.8509 million public employees excluding subsidised contracts. In 2024, 136,700 new direct pensions were granted: at a constant flow, five years would represent 683,500 retirements, and 521,000 would equal about 76% of that retirement-only flow. This is a capacity screen, not a forecast, because natural attrition includes other exits and occupations are not interchangeable.
The target must be rebuilt task by task: automatable, assistable, retained or reinforced; then identify positions truly avoidable, redeployable or left unfilled. The same employee cannot generate savings simultaneously under 2.09, 3.04, justice, health, anti-fraud and 9.07.
Productivity account only; €0 deficit effect by default. Fiscal conversion only through an owning documented non-replacement/removal line.
DINUM already plays an interministerial role in State digital policy; in 2026 DGAFP is steering an AI framework negotiation across all three civil-service branches, while ANSSI owns cybersecurity baselines. A new mission must therefore not recreate existing functions.
Prefer a lean steering team built on existing structures, with explicit mandate, staffing, budget and duration. Platform costs remain under 9.03 and training under 9.05.
Investment/OPEX to be costed; €0 direct saving.
Quarterly oversight is used to verify schedules, spending, incidents, measured benefits and safeguards. It is not itself a saving.
Reuse the same indicators as 9.11 and the same audit framework as 9.10 to avoid three parallel reporting chains.
€0; reporting/oversight costs may be isolated.
The measure must share the same audit mandate with 10.14, which already provides for a permanent annual audit mandate for the Court of Auditors.
Define one institutional responsible measure: a single annual audit can cover AI, costs, savings, risks, service quality and rights compliance.
€0; no additional saving.
The measure depends on defining non-gameable indicators: AI spending, unit cost, processed volumes, time saved, positions actually redeployed/unfilled, cash revenue actually collected, incidents, appeals and corrections.
Publish source data once and reuse it for Parliament, the Court of Auditors and the transparency platform; avoid duplicate reporting costs.
€0; indicator-production cost separated.
Current law already contains safeguards: GDPR Article 22 regulates solely automated decisions with legal or similarly significant effects, while the AI Act notably requires competent human oversight for high-risk systems; certain public bodies must also perform a fundamental-rights impact assessment.
The Delta-Sierra measure can become a simpler and broader rule than the legal minimum: no adverse decision is materialised without identifiable human validation, appeal route and logging. Review cost must be included in use cases.
€0 saving; human review time included in net project cost.
The platform should aggregate data already produced under 9.11 rather than create a parallel reporting system. It can publish use cases, suppliers, models, costs, evaluations, incidents, indicators and corrections, subject to protected secrets and cybersecurity.
A single data source feeds Parliament, the Court of Auditors and the public. Other Plan transparency platforms should reuse the same components rather than build a separate full system for every theme.
€0 saving; digital cost to be mutualised.
Public procurement already has national essential-data feeds: since 2024 procurement data are centralised on data.gouv.fr; in 2026 reporting thresholds remain €40,000 excluding tax under the normal regime and €25,000 under the simplified regime. This base makes large-scale statistical and algorithmic controls possible.
9.14 is the detector. The financial effect belongs to the line that materialises the action: 6.12/8.04 if a price is reduced, 11.12 if a suspicious contract is stopped or fraud recovered. An anomaly score is never a saving.
€0 direct; savings/revenue recorded under the owning line after evidence and cash collection/price reduction.
The French Constitution already sets a public-accounts balance objective in Article 34, but it does not directly prohibit an annual structural deficit. Germany still caps federal structural net borrowing at 0.35% of GDP, although the framework was amended in 2025, notably for Länder and certain defence expenditure.
10.01 is a governance rule. It cannot claim the savings of the measures it requires to be funded. The text must define scope, cycle, exceptions, amortisation, oversight and procedural consequences.
€0 direct; eligibility condition for the fiscal scenario.
The Swiss debt brake is embedded in Article 126 of the Constitution and sets an expenditure ceiling linked to cyclically adjusted revenue. It was approved by referendum on 2 December 2001 with 84.7% support. The precedent is a constitutional rule approved by referendum, not a mandatory referendum on every borrowing operation.
The Plan must specify what “referendum-backed” means in France: initial ratification of the rule, or popular authorisation for certain exceptions. The Swiss model must not be credited with a procedure it does not use.
€0 direct; institutional parameter and exception clause.
The 0.35% figure corresponds to Germany’s federal structural net-borrowing ceiling, but it is not exactly the same concept as the structural deficit of the French general-government sector. France’s current programming law sets a −0.4% potential-GDP medium-term objective, while the 2024 EU framework now uses the net-expenditure path as the annual operational indicator.
Precisely define the numerator, GDP reference, cyclical adjustment method and data revisions. Track the national ceiling alongside the EU net-expenditure path.
€0 direct; compliance gate.
Current law already contains a correction mechanism: LOLF Article 62 defines a significant deviation as 0.5% of GDP in one year or 0.25% per year on average over two years; the 2023-2027 programming law requires a return to the path within no more than two years. The automatic appropriation freeze proposed by 10.04 goes further.
The freeze must protect mandatory, sovereign and critical investment expenditure, include release/appeal rules, and cannot count as a saving until an appropriation is actually cancelled or left unspent.
€0 at trigger; saving only on actual net cancellations/non-spending.
The reduction in the number of MPs and senators has already been financially audited under Category 01. Measure 10.05 is the constitutional implementation vehicle, not a second headcount reduction to add.
Measure to which the saving is assigneds: 1.01 and 1.02. 10.05 remains €0 in the ledger.
€0; depends on 1.01/1.02.
This line materially overlaps measures 10.01 and 10.03. It should be retained as part of the constitutional package but does not create a second balance rule or a second saving.
Implementation line: it may be legally merged with 10.01/10.03 in the future constitutional bill.
€0 direct.
AI safeguards are already partly carried by GDPR, the AI Act and sectoral law. A constitutional amendment is therefore not required for every technical rule; it is justified only if a durable higher-level principle is to be entrenched, such as human control, transparency or fundamental rights.
10.07 is the higher-level legal vehicle; AI costs/gains remain in Category 09 and the relevant business areas.
€0 direct.
Public-pay transparency and caps are already addressed by 1.14 and 1.15. Much of the mechanism can be statutory or regulatory; constitutional entrenchment is justified only if the Plan wants the principle itself to be protected from ordinary-law reversal.
No additional gain under 10.08.
€0; refer to 1.14/1.15.
The constitutional levy cap is already measure 6.01. Measure 10.09 is its implementation in the constitutional package, not a second tax cut.
Any actual levy reduction remains costed and financed under Category 06.
€0; responsible measure 6.01.
Ministerial cabinet size is currently set by decree. Since 18 April 2026, the cap is 14 members for a full minister, with an exception of 19 for the Public Action and Accounts minister; delegated ministers are capped at 8, with an exception of 14 for Parliamentary Relations.
A cap of 10 can technically be set by decree. Constitutional entrenchment mainly prevents a future government from easily raising it. The financial saving remains owned by 1.08.
€0; responsible measure 1.08.
The CESE is directly embedded in Articles 69 to 71 of the Constitution. Abolition therefore does require constitutional amendment, followed by organic-law, asset, contract and workforce transition measures.
Net savings remain exclusively under 1.12.
€0; responsible measure 1.12.
Article 72 of the Constitution explicitly lists regions among territorial authorities. Full abolition therefore requires constitutional treatment and a mapping of transferred powers, staff, assets, debts and contracts.
No additional gain: measure to which the saving is assigneds are 2.01 for the administrative tier and 1.06/1.13 for the relevant regional elected/advisory bodies.
€0; refer to 2.01, 1.06 and 1.13.
Article 18 already allows the President to address Parliament convened in Congress; the statement may be debated in his absence, without a vote. The novelty in 10.13 is therefore a mandatory mid-term event, standardised content and auditable data.
A mere address does not require a new constitutional power. A binding obligation or follow-up vote could, however, require stronger legal grounding depending on the design.
€0 direct; marginal administrative cost to be shared with 10.15.
Article 47-2 already tasks the Court of Auditors with assisting Parliament and Government in financial execution oversight and policy evaluation. A specific annual mandate can therefore often be created by organic or ordinary law without creating a second audit institution.
Do not count the same saving twice with 9.10. Any savings uncovered by the audit belong to the corrected measures, not to the audit itself.
€0 direct; oversight cost to be documented.
Monitoring can be organised through statute and chamber rules. The financial objective is to avoid a new heavy structure: it should reuse 9.11 indicators, 10.14 audits and the fiscal register instead of creating parallel reporting chains.
A single data chain should feed Parliament, the Court of Auditors, HCFP and the public.
€0 direct; marginal secretariat cost to be shared.
Article 89 first requires identical adoption by both chambers. Amendment then becomes final through referendum; for a government constitutional bill, the President may instead choose Congress, requiring three-fifths of votes cast. A “referendum fallback” therefore cannot freely bypass identical bicameral approval.
The public page must distinguish government and parliamentary amendment bills, ratification referendum and any other referendum scenarios.
€0 direct.
A delivery unit can be created by decree. It should be designed as a lean implementation and consolidation team, without reproducing functions already handled by DINUM, DGAFP, DITP, the Budget Directorate, SGG or the 9.08 AI mission.
The unit cost is an implementation cost. Reform savings remain assigned to their measure to which the saving is assigneds.
OPEX to be documented; €0 own saving.
The Plan index states 5–8 billion euros/year, while the detailed section estimates only 650 million euros–1.3 billion euros/year for the cash cap. It also replaces part of the cash support with in-kind benefits through the citizen card, so a cash reduction is not a net saving until replacement benefits are costed.
Freeze the consolidated amount. Microsimulate household by household: cash removed − in-kind entitlements − card cost − induced take-up/non-take-up − transition costs. Never add 11.01 and 11.02 as two savings.
€0 by default pending microsimulation; retain 650 million euros–1.3 billion euros only as a historical gross scenario, not a net saving.
The detailed section proposes a card merging several benefit credentials and delivering in-kind rights for food, energy, transport and healthcare. Those rights carry real fiscal costs. A digital card may simplify administration but also creates system, cybersecurity, merchant-network and data-protection costs.
Treat 11.02 as the delivery mechanism for 11.01, not as a separate saving. Administrative gains belong to 9.07/2.04 only when an actual cost disappears.
CAPEX/OPEX to be documented; €0 direct saving.
Real-time monitoring may reduce errors, overpayments and delays, but the Plan also proposes automating CAF tasks; those productivity effects already sit inside the AI/civil-service perimeter.
Only amounts actually prevented or recovered and not already assigned to 6.10/11.05 may be booked. Staff time remains productivity until a cost is actually removed.
€0 direct; performance/fraud indicator.
Current law already contains residence-status conditions. For RSA, many foreign nationals must have held a work-authorising residence permit for at least five years, subject to statutory exceptions. The historical 6–8 billion euros estimate therefore cannot be built on the assumption that all benefits are available without conditions from arrival.
Audit RSA, housing, family benefits, healthcare and emergency support separately by residence status and applicable law. Subtract restrictions already in force and preserve constitutional, EU and treaty exceptions.
€0 by default; no saving before verified administrative base and legal review.
In 2025, nearly 3.1 billion euros of social fraud was “detected, prevented or reassessed”. Those categories are not equivalent to cash collected. The Family branch detected 509 million euros of fraud, but consolidation must track amounts actually prevented or recovered net.
Create one 6.10/11.05 ledger separating detected, assessed, prevented, collected, control costs and reversals. 6.10 is the AI tool; 11.05 is the substantive programme. Each euro can belong to only one line.
Net revenue/saving only for incremental amounts actually prevented or collected above the 2025 baseline.
Mutualised funding exists in the Labour Code. For one mission of the joint fund, the allocated amount cannot be below 73 million euros. The historical 1.2 billion euros figure is therefore much broader than this channel and may include the tax credit, union release time or other resources already isolated in 11.07/11.08.
Build a channel-by-channel ledger: joint fund, grants, secondments, premises, union time and tax expenditure. Exclude 11.07 and 11.08 from the amount owned by 11.06.
€0 by default; retain 1.2 billion euros as a source figure requiring decomposition.
The trade-union tax credit remains in force: 66% of dues, capped at 1% of the relevant gross income. The Plan’s 144 million euros figure is historical and must be reconciled with the latest tax-expenditure statement before consolidation.
Book only the verified current tax-expenditure cost. Remove any amount already counted under 11.06, which must not include the same tax expenditure in its 1.2 billion euros figure.
144 million euros = historical reference; active amount only after budget refresh.
The historical 800 million euros figure must be rebuilt across all three civil-service branches using actual union release-time usage, loaded staff cost and continuing social-dialogue duties. One hour of reduced release time is not automatically a salary saving.
Separate time returned to service, avoided position, avoided replacement and payroll actually removed. Exclude any amount already included in 11.06 and in productivity lines 2.04/9.07.
€0 by default; 800 million euros is a source figure to rebuild.
Current law already requires several large public employers to publish annually the aggregate amount of their ten highest remunerations and the gender split. 11.09 goes much further by seeking “full publication” and must therefore define scope, granularity, privacy and public-interest safeguards.
No direct fiscal gain. Pay anomalies revealed by publication belong to the relevant remuneration reforms, not to the transparency platform.
€0 direct; publication cost to be documented.
In 2026, essential procurement-data reporting remains at €40,000 excluding tax under the standard regime and €25,000 under the simplified regime. 11.10 therefore does not invent transparency at €25,000; it would extend/standardise automatic publication and data depth below the normal threshold.
Measure compliance cost and actual coverage. Price savings or fraud findings belong to 6.12/8.04/11.12, not to 11.10.
€0 direct; marginal OPEX to be documented.
Interest declarations already exist for certain jobs whose level or functions justify them. The Civil Service Code provides for confidentiality in the staff file except for authorised access. General publication for all category A/A+ staff would therefore change both scope and publicity, not merely publish already-public data.
Precisely define covered jobs, publishable interests, redactions, duration and appeals; apply privacy/proportionality review. Potential savings from avoided conflicts are not budgeted ex ante.
€0 direct; HATVP/administrative cost to be documented.
Preventive control of suspicious contracts directly overlaps the anomaly detector in 9.14 and the margin reform in 6.12/8.04. Procurement open data under 11.10 is also an input to that control.
Use one procurement ledger: anomaly → review → contract cancelled/renegotiated → net saving. No euro may be assigned simultaneously to 9.14, 11.12 and 6.12/8.04.
€0 by default; saving recognised ex post on contracts.
Current law already provides, for persons covered by the 2013 transparency act, up to three years’ imprisonment and a €45,000 fine for failure to file, substantial omission or misleading valuation in specified declarations. 11.13 must therefore define what is strengthened: scope, intent requirement, penalty or enforcement.
Fines are not a structural saving and must not finance the Plan. Do not count the same saving twice with the wider filing duties in 11.11 and enforcement capacity in 11.14.
€0 budget financing; fine revenue excluded from the prudent baseline.
The HATVP is already an independent authority with an annual budget. Its website reports 79 FTE in 2025 and a 2024 initial budget of 12.3 million euros in payment appropriations. The 2026 budget bill requested 11.53 million euros in payment appropriations, while the Senate also noted unmet needs linked to new missions. correction: these needs must not be extrapolated at constant human productivity. An AI architecture can pre-screen filings, cross-match registers and databases, flag inconsistencies and prepare control memoranda.
Size the organisation only after a task audit: delete / automate / assist / human-only. Humans retain adversarial review, conflict assessment, sensitive follow-up and decisions. AI productivity remains free of double counting from 9.07; savings uncovered by HATVP stay assigned to the corrected policies.
AI OPEX + human review + infrastructure − avoided/redeployable workload. €0 autonomous saving until an actual cost is removed.
The 2023 reform already closed several major schemes to new entrants, while SNCF had been closed since 2020. The Senate nevertheless stresses that closure does not erase pensions or accrued rights: in 2026 the Special Social and Pension Schemes mission remains close to 6 billion euros, 69% of which goes to SNCF and RATP, both already closed.
Start with an exhaustive census of schemes still open. Do not book the disappearance of any balancing subsidy merely because a scheme closes to new entrants. Parameter-related savings belong to 12.02.
€0 short-term by default; separate long-term actuarial effect.
The Plan states 400–800 million euros/year at the five-year horizon, 1.5–2.5 billion euros/year at ten years and 5–6 billion euros/year in the very long run. The roughly 6 billion euros mission baseline, however, finances accrued rights and demographic imbalances; it is not a base that can be cut by applying a percentage.
Build a cohort-by-cohort actuarial model: age, service, accrued rights, retirement date, current rule, convergence path, contributions and compensation. 12.01 receives no second saving.
Retain 400–800 million euros/year as the source five-year scenario, with status “actuarial consolidation required”.
Budget documents already publish appropriations and subsidies for several schemes, but the perimeter remains fragmented across budget missions, assigned taxes, compensation mechanisms and other flows.
Publish an annual scheme ledger: contributions, balancing subsidy, assigned taxes, compensation, contributors/pensioners, cost per pensioner and liabilities. Any savings uncovered remain assigned to 12.02 or the substantive measure concerned.
€0 direct; marginal publication cost to document.
Article 34 of the Constitution already assigns the fundamental principles of social security to statute. A constitutional ban on new derogatory schemes would add a safeguard, but its scope must be precisely defined and objectively justified occupational differences or public-interest distinctions preserved.
No autonomous saving. Define “derogatory scheme”, admissible exceptions and proportionality review; never duplicate the savings of 12.01/12.02 into 12.04.
€0 direct.
The CRE confirms in 2026 that the employee tariff remains in force and treats its costs in network regulation; for RTE, the amount retained for 2025 is 66 million euros and volumes remain incentive-regulated. The Plan also cites a cost above 700 million euros in 2024 for the EDF group. Perimeters must therefore be aligned before costing.
The volume cap owns the reduction in subsidised consumption. Calculate beneficiaries × consumption × price gap, then separate EDF, other IEG firms and regulated networks. The gain first improves enterprise accounts; it reaches the State budget only through taxes, dividends or lower financing needs.
250–350 million euros/year = source scenario, to verify against total cost and consumption data.
The IEG employee-tariff scheme covers beneficiaries and, depending on circumstances, may extend beyond the primary residence. The Plan historically estimates secondary-home exclusion at 40–70 million euros/year but does not yet provide a verified contemporary count of contracts and volumes concerned.
Count only identified secondary delivery points and their cost gap. If the 12.05 cap already removes the same volumes, calculate 12.06 on the residual, not the gross base.
40–70 million euros/year = source assumption; €0 consolidated until the base is verified.
The Plan proposes restoring standard subscription charges and taxes. Contemporary energy tariffs combine supply, network and tax components; CRE regulation also treats price and tax effects within employee-tariff costs. The historical 60–90 million euros therefore cannot be booked entirely as State revenue.
Split line by line: subscription/supply to supplier or network operator; excise/VAT/CTA and other levies to the beneficiary authority. Count only net incremental revenue after interactions with 12.05/12.08.
60–90 million euros/year = mixed source envelope; not consolidated before tax/accounting allocation.
CRE frameworks confirm that retirees with at least fifteen years in the IEG sector can benefit from the preferential tariff. The Plan proposes ten-year convergence, with 150–250 million euros/year at maturity. However, this population is already partly affected by the volume cap, secondary-home exclusion and subscription/tax measures.
Apply 12.08 last to the residual retiree/beneficiary cost after 12.05–12.07. Never add the four gross ranges as if their bases were disjoint.
150–250 million euros/year = ten-year source scenario; residualise before consolidation.
The public dossier remains self-contained and sourced. These two books develop the full State-reform architecture and the use of artificial intelligence to transform public services.

Public finances, institutions, implementation calendar and reform measures.
View the book →
How artificial intelligence can simplify, accelerate and improve public services.
View the book →