French debt: holding the rating agencies to account

Moody’s concluded a periodic review of France on October 23, 2026, leaving its Aa3 rating and negative outlook unchanged. AFP distinguishes the publication from a rating action. S27 The finance ministry portrayed the outcome as recognition of the economy’s strength and the credibility of its budget. That political reception leaves a more demanding question unanswered: how much room does the rating allow if the expected improvement takes longer to arrive? S01 S02

Reuters reports that Moody’s assessment assumes, among other things, a deficit-reducing budget for 2027 and some easing of financing costs and social tensions in the following months. Fiscal improvement still has to be delivered. Easier funding is an expectation. A rating built partly on that outlook should make its vulnerability to disappointment intelligible. S02 S11

France’s creditworthiness deserves scrutiny. So does the institution assessing it. The experience of 2008 makes that second obligation unavoidable. Ratings sold as independent judgments helped risks spread when those risks were poorly measured. Moody’s 2017 civil settlement documented departures from its published methods. That record warrants close examination of current practices, without predetermining the correct rating of every borrower. S19 S20

Public documents allow a comparison between the assumptions attributed to Moody’s and France’s actual financing requirements. They also raise a wider question, increasingly relevant to artificial intelligence: how should a rating account for incomplete information, promised revenue and shared dependencies? This is a question about analytical standards, irrespective of the borrower’s nationality.

A forecast needs a visible margin for being wrong

Credit is repaid in the future. Assessing a borrower today necessarily requires assumptions about revenue, spending, economic conditions and access to financing. Rating agencies describe their judgments as forward-looking credit opinions. Removing assumptions would exclude the very period during which the lender is waiting to be paid. S18

The quality of that judgment depends on how it handles adverse developments. For France, a reader should be able to understand what a weaker budget, disappointing growth or persistently high yields would do to the assessment. Those developments can reinforce one another: weaker activity reduces receipts while more expensive refinancing adds to expenditure. Testing each disappointment in isolation can miss their combined force. This is our proposed test of the reported assumptions, rather than a reconstruction of an internal Moody’s model.

There is an important documentary boundary. We did not obtain Moody’s complete original statement. Its assumptions are reported through Reuters. We therefore cannot certify the severity of the agency’s downside tests or assert that it performed none. The critical demand is for inspectable sensitivities, review triggers and reasons for keeping the rating unchanged. The ministry’s reassuring interpretation cannot supply that evidence. S01 S02

AFP reports that Moody’s forecasts a 5.2% deficit in 2027, above the government’s 5% target. The agency also argues that France’s debt structure cushions higher rates initially, while a prolonged increase would eventually weaken debt sustainability. These reservations belong in any examination of its assumptions. S27

The negative outlook itself qualifies the assessment. Keeping the rating unchanged does not, by itself, endorse the government’s programme. The projected 2027 general-government deficit of 5% of GDP still depends on legislation and execution. The budget figures retain their original dates and sources, distinguishing the AFT release from the submitted bill. S01 S11

A second distinction keeps the criticism properly targeted. A bond yield reflects more than default risk, including monetary conditions, maturity and liquidity. Rising yields alone cannot establish that a credit rating is wrong. They do, however, require a fresh assessment of how quickly more expensive financing can erode the borrower’s room for manoeuvre. Treasury data offer a practical way to test that judgment.

A market reference is only one of the relevant rates

The AFT’s daily ten-year reference, the TEC 10, stood at 5.05% on October 23, after 5.13% the previous day. A week earlier it had been 4.79%. That is a rise of 26 basis points, or 0.26 percentage points. The observations provide no reading after Moody’s evening announcement. Attributing Friday’s decline to that announcement would reverse the order of events. S04 S02

A second measure describes borrowing undertaken over the course of the year. In its October bulletin, the AFT reports a weighted average yield of 3.64% on medium- and long-term issuance through October 19, against 3.14% for the whole of 2025. That average combines different issuance dates and maturities. Subtracting it from the October 23 TEC 10 would not produce a meaningful risk premium. It does, however, show that higher financing costs are already appearing in actual issuance. S05

The cost of the outstanding portfolio follows another timetable. At September 30, the average remaining life of the State’s marketable debt was eight years and 158 days. Existing fixed-coupon bonds keep their contractual payments until redemption. New borrowing conditions therefore feed into the portfolio mainly as maturing debt is replaced or additional funds are raised. Inflation-linked securities have a separate adjustment mechanism. S06 S09 S14

An average maturity buys time, but it conceals very different repayment dates. Short-term Treasury bills, known as BTFs, accounted for €217.658 billion at the end of September, approximately 7.5% of outstanding marketable debt. Their average remaining life was 129 days. Part of the portfolio returns to the market quickly, while some bonds remain outstanding for decades. S06 M1

How a 1.25% coupon can cost the issuer 5.06%

The May 2038 OAT, a French government bond, provides a concrete example. The Treasury reopened an existing bond: the additional securities have the same coupon and redemption date as those already outstanding. Reopening builds a larger, more readily tradable line instead of creating numerous small issues. The new tranche’s price adjusts to the demand available at the auction. S07 S08 S09

For each €100 of face value, the investor is due €1.25 a year and the principal at maturity. A purchase price of €67.10 before accrued interest also leaves €32.90 between the clean purchase price and final principal repayment. The reported yield to maturity of 5.06% takes the payments and their timing into account. The €32.90 is spread across the remaining life of the investment; it is not an annual interest payment. S07 S08

That distinction matters on the issuer’s side. The competitive allocation for this bond was €2.134 billion of face value. Using the rounded average price reported by the AFT, it raised approximately €1.432 billion before accrued interest. The roughly €702 million gap to face value belongs to the remaining financing period. Treating it as an immediate realized loss would misrepresent the transaction. Subsequent non-competitive bids brought the final tranche to €2.828 billion; those two volumes refer to different stages of the issuance process. S07 S08 M1

The Treasury could instead issue a bond with a coupon closer to the prevailing market yield. Its price would then be closer to par, meaning €100 paid for €100 of principal. That choice changes the balance between coupon payments and the initial issue price. Under comparable market conditions, an older bond’s small coupon offers no economic shortcut. A parliamentary report on the 2024 accounts explains why issue premiums and discounts accompany reopenings and why different accounting frameworks recognize them differently. S09 S13

A smaller deficit still leaves a large refinancing requirement

The initial 2027 proposal puts the State’s financing requirement at €339.7 billion, compared with an updated forecast of €311.7 billion for 2026. The AFT attributes most of that €28 billion increase to a €19.4 billion rise in medium- and long-term debt redemptions. It also points to the need to replace European financing for the national recovery and resilience plan as that funding reaches its scheduled end. S10

Two uses of the proceeds explain how a smaller deficit can coexist with greater market borrowing. Some new money replaces principal returned to investors whose bonds have matured. Another portion covers the gap between the year’s spending and revenue. The first requirement follows the maturity schedule inherited from earlier issues. The second is shaped by fiscal decisions and economic activity, among other factors.

The draft bill allocates €189.2 billion to medium- and long-term redemptions, including inflation adjustments. The State budget deficit to be financed is €156.4 billion. Adding €0.6 billion for debt taken over from SNCF Réseau and applying a cash adjustment of minus €6.5 billion produces the €339.7 billion total. The State deficit in euros and the general-government deficit target as a percentage of GDP cover different institutional boundaries: the latter also includes other parts of the public sector. S11

The proposed issuance programme calls for €340 billion of medium- and long-term securities net of buybacks. The qualification is important. The AFT can issue a larger gross amount while using some proceeds to repurchase existing bonds before they mature. The table also projects a €2.2 billion net increase in outstanding BTFs. That is a change in the stock at year-end. Rolling over short-term bills can require much more issuance during the year. S10 S11

The 2027 maturity schedule is already changing

The October bulletin updates the maturity reconstruction examined in l0g’s earlier French debt coverage. The four conventional fixed-rate OATs maturing in 2027 total €157.004 billion at September 30, down from €165.642 billion a month earlier. The €8.638 billion decline is a reminder that a maturity schedule is a dated snapshot of outstanding contracts. Debt-management operations can change it before repayment falls due. S05 S17 M1

A fifth line, the inflation-linked OAT maturing in July 2027, has a face value of €21.737 billion. Including inflation accumulated through the end of September brings its outstanding amount to €30.662 billion. Across all five bonds, the snapshot totals €187.666 billion. That is close to the €189.2 billion of redemptions forecast in the budget bill, but the two amounts have different reference dates and assumptions. Future inflation and debt-management operations can still alter the amount ultimately repaid. S05 S11 M1

The comparison identifies the scale of the task. Substantial principal will need replacement funding even if the annual deficit falls. The cost will depend on the maturities of the new securities, the prices obtained and when they are issued. In the documents reviewed, the AFT had yet to publish the detailed 2027 financing programme; it was scheduled for December. Applying a single ten-year yield to every forthcoming redemption would erase those choices. S10

Inflation enters the budget and the cash account at different times

An inflation-linked bond’s principal increases with the price index specified in its contract. The associated increase can enter the budgetary debt charge before the Treasury pays the additional principal to the holder. When the bond matures, the cash repayment can therefore include inflation accumulated over several years. S14 S11

The 2027 bill makes the distinction explicit. It includes an €8.5 billion annual indexation provision in the budget deficit, without a corresponding cash payment for that provision, and €9.7 billion of inflation uplift payable on securities maturing during the year. The former tracks the charge assigned to the financial year; the latter follows the contracts being redeemed. Treating them as rival estimates of the same expense would obscure how the funding requirement is calculated. S11

To move from the budget deficit to the cash requirement, the financing table removes the €8.5 billion provision. It also adds €2 billion of payments under the future-investment programmes. Together, those adjustments explain the minus €6.5 billion under “other cash requirements.” The inflation uplift actually repaid is already included in the €189.2 billion of redemptions. The reconciliation prevents a charge recognized at a different date from generating a duplicated financing requirement. S11 M1

Issue discounts require a related adjustment between face value and cash received. For 2027, the government assumes €8 billion of net issue discounts. Other cash resources total minus €5 billion after €3 billion of indexation supplements received when issuing inflation-linked securities. Readers need not memorize each accounting line. The useful principle is that a euro of issued principal, a euro of budgetary interest and a euro available in the Treasury’s account can arrive at different times. S10 S11

The bill rises before the entire portfolio rolls over

In its September 29 release, the AFT puts the budgetary debt charge at €59.3 billion in the initial 2026 budget, revised to €62.6 billion, followed by €72.9 billion proposed for 2027. Within that publication, the first change is a €3.3 billion revision for 2026; the second is a €10.3 billion increase between its two annual forecasts. S10 M1

The budget bill submitted on October 1 presents a different update: its explanatory statement gives a €4.1 billion upward revision for 2026, followed by a €9.6 billion increase in 2027 against its revised 2026 base. Proposed appropriations for programmes 117 and 355 total €72.934 billion in 2027. A detailed reconciliation of its updated 2026 base with the earlier release remains unavailable in the reviewed documents. The chart therefore retains the explicitly dated September 29 AFT snapshot. S11

These amounts bring together mechanisms operating at different speeds: coupons on outstanding securities, bill rollovers, refinancing and indexation. Their proximity to the recent move above 5% does not establish that the week of October 19 caused the full increase in the projected debt charge. The budget scenario was published before that market movement. S06 S10 S11

Gradual transmission has an important consequence. Even if market yields decline, particularly cheap legacy securities may continue to mature and be replaced by more expensive borrowing. The debt charge can then rise while current yields fall. Conversely, a market-rate increase works through the fixed-rate portfolio over time rather than immediately resetting every coupon. Assessing the exposure requires both the rollover schedule and the terms of the securities entering the portfolio.

Non-resident ownership and quarterly measurement

Non-residents held 59.66% of marketable State debt in the second quarter of 2026, compared with 49.82% at the end of 2022, according to Banque de France data reproduced by the AFT. The calculation uses market values. It classifies holders by residence, without identifying their nationality or investment strategy. A fund based elsewhere in the euro area also falls within the non-resident category. S05 S15

The observation date and valuation basis rule out two common shortcuts. This quarterly figure cannot reveal purchases or sales made in October. Nor can its market-value share simply be multiplied by the €2,896.181 billion nominal stock measured at the end of September to manufacture an amount of “foreign debt.” Market prices, portfolio composition and transactions can all affect the reported shares. S06 S15

A diversified investor base broadens access to funding. It also exposes the issuer to international investors’ comparisons between competing bonds. A closer assessment would require information on flows and on how holders fund their positions, including any use of leverage. The non-resident category cannot measure that exposure by itself. An investor’s location says little about whether it can keep holding a security through a period of market stress.

Auctions offer more immediate evidence of demand. On October 15, bids for the four securities offered amounted to between 2.23 and 2.32 times the competitive allocations, which totalled €12 billion. Buyers were present at the prices and yields available that day. Those bids are neither a cash reserve the government can draw on later nor a commitment to participate in the next auction. S12

Real protections still need to be tested

France retained market access in the operations examined here. By October 19, the AFT reported completion of 92.31% of its annual medium- and long-term issuance programme net of buybacks. The maturity profile also slows the transmission of new yields to fixed-coupon debt. These are substantive reasons to distinguish pressure on the public finances from imminent insolvency. S05 S06

The ECB’s Transmission Protection Instrument can address unwarranted, disorderly market dynamics that threaten monetary-policy transmission. Its framework requires an assessment of fiscal policies, debt sustainability and other conditions. A 5% yield triggers no automatic purchases, and the framework document establishes no decision to intervene for France. Conditional protection must remain conditional in a credit assessment. S16

Those buffers justify looking beyond a bad trading session. They also require the agency to explain how long they remain sufficient. The relevant test compares them with repayments, interest expenses and revenue, then identifies the assumptions that would change the judgment. France’s financing data provide substantial material for that exercise. They cannot produce a replacement credit rating from a handful of selected figures.

The responsibility established by 2008

In its majority conclusions published in January 2011, the US Financial Crisis Inquiry Commission assigned rating agencies an essential role in the sale of mortgage securities and in the disruption that followed their downgrades. The commission also identified failures elsewhere in the system. Agencies’ central responsibility belongs in that chain alongside lenders, intermediaries, investors and supervisors. S19

One finding is particularly stark: 83% of the mortgage securities Moody’s rated AAA in 2006 were subsequently downgraded. The denominator is that cohort of securities. The percentage measures rating revisions, rather than defaults, investor losses or dollar exposure. S19

Mortgage repayment depends on households, incomes and housing conditions. When many securities rely on those same conditions, apparent diversity can conceal a shared weakness. In July 2008, the Committee on the Global Financial System highlighted model limitations and called for disclosure of ratings’ sensitivity to default, recovery and correlation assumptions. A credible test must examine risks that deteriorate together, rather than assume convenient independence. S21

The record goes beyond mistaken forecasting. In January 2017, Moody’s reached a civil settlement of nearly $864 million with US authorities over ratings of mortgage securities and collateralised debt obligations, which pool debt claims. Among the facts acknowledged was its use, beginning in 2004, of more lenient standards than those it had published for certain Aaa-rated CDOs, without informing the broader market. The conduct therefore concerned adherence to announced methods as well as the eventual outcome of a forecast. The settlement was not a criminal conviction. S20

That history supports a concrete present-day requirement: an agency’s methods, their application and any necessary corrections should be open to meaningful scrutiny. The historical findings establish no manipulation of France’s 2026 rating. They do make it difficult to argue that the agencies’ reputations alone should settle doubts about analytical quality.

The borrower can also be the evaluator’s customer

The business model adds a lasting tension. S&P continues to describe an issuer-pays model, arguing in part that it supports public access to ratings. The agency sits between a customer seeking funding and investors expecting an independent assessment. That structural conflict can exist without evidence of improper commercial interference in a particular decision. S18

Moody’s 2017 settlement gives the concern a documented historical form. It acknowledged leaving business-unit managers to resolve the tension between rating quality and winning business from the issuers selecting the agency. The finding explains why the separation of sales and analytical work deserves examination. Our evidence contains neither the France–Moody’s contract nor its invoice. It cannot establish any specific payment or commercial pressure in the current sovereign case. S20

Safeguards feature in the agencies’ stated arrangements. S&P describes independent methodology checks, separation of commercial and analytical functions, and analyst compensation that is not tied to assigned ratings. Moody’s accepted compliance measures in the 2017 settlement. Such provisions matter, while their effectiveness in an individual case requires separate verification. S23 S20

The influence extends beyond the commercial relationship. Where an investment rule or contract links an automatic consequence to a rating threshold, a downgrade can trigger simultaneous decisions by several investors. In October 2010, the Financial Stability Board called for less mechanistic reliance on ratings to reduce cliff effects and herding. This is a conditional explanation of ratings’ power, not evidence of forced sales of French debt in October 2026. S22

A private opinion can become a collective rule of action. That transformation increases the responsibility of the agency, and of the regulators and investors who choose to rely on it. An external rating should not displace the judgment of the institution committing the money.

European supervision of rating agencies

In the European Union, the European Securities and Markets Authority, ESMA, directly supervises credit rating agencies. It reviews regulatory submissions and complaints and conducts investigations, which may include on-site inspections. Breaches can lead to fines or withdrawal of registration. Regulatory compliance and the merits of an agency’s credit judgment remain separate questions. S26

AI financing is already a contemporary test

On October 19, Reuters reported concerns raised by Mark Habib, S&P’s global head of digital infrastructure, about incomplete visibility into some AI-linked commitments. He considered the ratings consistent with the agency’s long-term approach at that time. S&P had downgraded Oracle to BBB− in July, according to the same report. The criticism therefore concerns the treatment of uncertainty and the criteria for changing a judgment. S24 S25

Acknowledging a blind spot is useful. The next question is how it changes the assessment: what safety margin covers poorly documented commitments, what further information is needed, and what finding would trigger a review? These are questions an evaluator should answer. We do not have a complete set of answers from which to recalculate the ratings.

The test can be illustrated without inventing a company’s accounts. Consider a project relying on a customer that both reserves computing capacity and guarantees its financing. If that customer needs less capacity, operating revenue could weaken just as the guarantee becomes more important. Its legal scope, duration and the guarantor’s financial resources would then need examination. Contractual protection can be effective; treating each promise as an entirely independent safeguard requires justification.

The borrower, the project company owning the infrastructure and the specific security being rated must also be distinguished. Group support, priority claims on assets and contractual restrictions can change the risk of a particular bond. For equipment that might be resold, the relevant recovery value is what a stressed sale could deliver after costs and delays. A model’s several decimal places cannot turn an uncertain resale price into a contractual payment. S18 S21

The same analytical discipline applies to states and companies, though their resources and legal obligations differ. France’s assessment needs tests of fiscal improvement and refinancing costs. An AI project’s needs tests of revenue, counterparties and guarantees. In both cases, weak assumptions should remain visible in the judgment delivered to the reader.

Accountability after France’s rating review

The evidence establishes higher issuance costs, substantial forthcoming repayments and a rising projected budgetary debt charge. It also documents the State’s buffers. Those facts support demanding scrutiny of the rating, without mechanically identifying the grade Moody’s should have assigned. S05 S10 S11

The agency should explain why its assumptions are sufficiently conservative, what happens when they fail and when it would change its judgment. The government should demonstrate how its commitments reach the accounts. Celebrating an unchanged rating satisfies neither obligation.

After the failures documented around 2008, confidence must rest on methods that can be checked, conflicts that are effectively controlled and errors that are examined. Those requirements apply to French sovereign debt and AI financing alike. Taxpayers and creditors need to understand the promises supporting their financial exposure.

Scope of the investigation

Revised and extended on October 25, 2026. Market observations end on October 23, average issuance yields on October 19, debt stocks on September 30 and holdings data in the second quarter. Budget figures refer to the initial 2027 bill. No interviews or confidential rating files were obtained.

The financial data have been checked against official publications. Successive budget-charge presentations retain their own dates and comparison bases. Moody’s specific assumptions are attributed to Reuters and AFP because its complete original statement was not obtained. The 83% figure comes from the US commission’s majority conclusions. S&P’s current compliance arrangements remain the agency’s own description. The AI financing example is illustrative and does not describe any identified contract.

For related mechanisms, our French debt refinancing analysis follows maturities and coupons, the Citi facility for Nebius examines customer commitments and AI debt, and circular AI financing connects clients, suppliers and financiers.

Sources and evidence

1. [S01] Ministère de l’Économie. Notation Moody’s : réaction de Roland Lescure.
   Source document
   Interested primary government source. Confirms Aa3 and the negative outlook; budget credibility is the ministry’s interpretation.
2. [S02] Reuters / Boursorama. Moody’s maintient la note souveraine de la France à Aa3.
   Source document
   Secondary access to Moody’s assumptions. Reuters publication October 24 at 00:29; underlying event October 23. The complete original agency statement was not obtained.
3. [S03] Reuters, syndicated by Euronext. France spared Moody’s ratings downgrade as it struggles with unrest. Reuters dispatch dated October 23; Euronext displays October 24 at 02:23, with no verified time zone. Same reporting as S02. Reuters report
4. [S04] Agence France Trésor. TEC 10 du jour.
   Source document
   Daily market reference. Sixteen observations from October 1 to 23, 2026; no October 5 value in the reviewed table. No imputation and no whole-portfolio borrowing-cost interpretation.
5. [S05] Agence France Trésor. Bulletin mensuel n° 437, octobre 2026.
   Source document
   October 2026 bulletin, precise publication day not established. Page 1: issuance data through October 19. Page 3: holdings in Q2 and stock at September 30. Page 5: September 30 security-level amounts. The inconsistent net-issuance series on page 2 is excluded.
6. [S06] Agence France Trésor. Encours de la dette négociable.
   Source document
   Updated October 1, covering September 30, 2026. Marketable State debt, not Maastricht general-government debt.
7. [S07] Agence France Trésor. Adjudication du 1er octobre : 11,999 milliards d’euros d’OAT long terme.
   Source document
   Competitive auction October 1, settlement October 5, 2026. Later non-competitive allotments raise the tranche: €2.134bn and €2.828bn refer to different stages.
8. [S08] Agence France Trésor. OAT 1,25 % 25 mai 2038, FR0014009O62.
   Source document
   Dynamic security record. Contract terms and final issuance history, including non-competitive allotments. The reported yield is not recalculated.
9. [S09] Agence France Trésor. Présentation des OAT.
   Source document
   General documentation on reopening, clean prices and coupons. Publication date not established; checked October 25, 2026.
10. [S10] Agence France Trésor. Besoins et ressources de financement de l’État en 2027 et point sur 2026.
    Source document
    September 29, 2026. Updated 2026 forecasts and proposed 2027 funding. The 3.55% issuance average in this release is superseded by 3.64% in the October bulletin.
11. [S11] Gouvernement / Assemblée nationale. Projet de loi de finances pour 2027, n° 3210.
    Source document
    Submitted October 1, 2026. Initial draft, not enacted legislation. Article 57, funding table and explanatory notes; relevant debt-charge annexes.
12. [S12] Agence France Trésor. Adjudication du 15 octobre : 12 milliards d’euros d’OAT moyen terme.
    Source document
    Competitive auction October 15, settlement October 19, 2026. Bids at one auction do not secure demand for future issues.
13. [S13] Assemblée nationale, commission des finances. Rapport n° 1492, tome I, sur les comptes de 2024.
    Source document
    Registered June 4, 2025, covering the 2024 accounts. Used only for the accounting mechanism, never as 2026 results or forecasts.
14. [S14] Agence France Trésor. Présentation des OAT€i.
    Source document
    General terms for euro-area inflation-linked OATs. The real coupon applies to indexed principal. Publication date not established; checked October 25, 2026.
15. [S15] Agence France Trésor / Banque de France. Principaux chiffres de la dette : détenteurs.
    Source document
    Banque de France data reproduced by AFT. Q2 2026 holdings measured at market value; classification by residence. Rounded category shares total 100.1%. No inference about October flows or nationality.
16. [S16] Banque centrale européenne. The Transmission Protection Instrument.
    Source document
    Framework published July 21, 2022. Activation remains conditional and discretionary. This document establishes neither France’s current eligibility nor a future intervention.
17. [S17] Agence France Trésor. Bulletin mensuel n° 436, septembre 2026.
    Source document
    September 2026 bulletin, precise publication day not established. Page 5 snapshot at August 31, used solely for comparison with the newer September data.
18. [S18] S&P Global Ratings. Understanding Credit Ratings.
    Source document
    Interested agency source. Explains forward-looking credit opinions and the issuer-pays model. It does not disclose the France–Moody’s contract.
19. [S19] Financial Crisis Inquiry Commission. The Financial Crisis Inquiry Report.
    Source document
    Majority findings, printed p. xxv / PDF page 25, checked in the report. The 83% figure concerns subsequent downgrades in the 2006 Moody’s AAA mortgage-security cohort, not defaults, losses or dollars outstanding.
20. [S20] U.S. Department of Justice. Justice Department and State Partners Secure Nearly $864 Million Settlement With Moody’s.
    Source document
    Civil settlement of nearly $864m, announced January 13, 2017, with factual acknowledgments and compliance undertakings. Not a criminal conviction or evidence about France’s current rating.
21. [S21] Committee on the Global Financial System / BIS. Ratings in structured finance: what went wrong and what can be done to address shortcomings?.
    Source document
    July 2008 report, website dated July 4. Printed pp. 1 and 13 / PDF pages 7 and 19. Supports scenario sensitivity and correlated-risk scrutiny; it is not an assessment of today’s AI financing.
22. [S22] Financial Stability Board. Principles for Reducing Reliance on CRA Ratings.
    Source document
    October 27, 2010 principles. Rating thresholds can trigger herding and cliff effects where rules or contracts hard-wire them. No such event is established here for France.
23. [S23] S&P Global Ratings. Transparency Statement.
    Source document
    Interested corporate statement describing independent methodology checks and separation of analytical and commercial functions. Not independent verification of compliance on any specific rating.
24. [S24] Reuters. Top hyperscalers likely to have more AI-linked debt than now known, S&P Global warns.
    Source document
    October 19, 2026. Secondary report of S&P’s remarks on incomplete visibility. Does not establish an amount of hidden liabilities, fraud or a misrating.
25. [S25] S&P Global Ratings. Oracle Corp. Downgraded To BBB-/A-3 From BBB/A.
    Source document
    Only the public downgrade-notice title was accessible. July 2026 timing comes from Reuters; the full rationale was not obtained. This is a historical action, not a claim about every subsequent rating action.
26. [S26] ESMA, Credit Rating Agencies. Direct EU supervision, desk-based reviews, investigations and enforcement powers. Regulatory source checked October 25, 2026. Official document
27. [S27] AFP, syndicated by TV5Monde. Moody’s laisse la situation s’éclaircir avant de se prononcer sur la note française. Updated October 23, 2026 at 22:22; periodic review, 5.2% deficit forecast for 2027 and the cushioning role of the debt structure. Separate reporting from Reuters; complete Moody’s text not obtained. AFP report