Primary documents

1. International Monetary Fund, Global Financial Stability Report, October 2026, Chapter 1, “Beneath the Resilience: Rising Risks, Unresolved Vulnerabilities,” published 13 October 2026. https://www.elibrary.imf.org/display/book/9798229055673/CH001.xml — retrieved and read in full; the captured text is 105,786 characters.

2. International Monetary Fund, Global Financial Stability Report, October 2026, abstract and excerpt. https://www.elibrary.imf.org/display/book/9798229055673/9798229055673.xml

3. Board of Governors of the Federal Reserve System, Financial Stability Report, May 2026, overview page (asset-valuation summary). overview — HTTP 200 observed on 1 November 2026.

4. Board of Governors of the Federal Reserve System, Financial Stability Report, May 2026, Section 1, Asset Valuations, data as of 23 April 2026. asset valuations — HTTP 200 observed on 1 November 2026.

5. Board of Governors of the Federal Reserve System, Financial Stability Report, November 2025, Section 1 and Box 1.1, “Artificial Intelligence and Algorithmic Trading.” asset valuations — HTTP 200 observed on 1 November 2026.

6. Nassira Abbas, Charles Cohen, Dirk Jan Grolleman and Benjamin Mosk, “Artificial Intelligence Can Make Markets More Efficient—and More Volatile,” IMF Blog, published 15 October 2024. imf.org

7. Canadian Securities Administrators, 2025 Systemic Risk Committee Annual Report, released 12 February 2026 via the Ontario Securities Commission. osc.ca — retained for context; not quoted in the report.

Quotations relied on in the report

Each span below was copied verbatim from the retrieved text and re-tested mechanically against that text. All 36 quotation and figure spans resolved exactly. The spans are grouped by the document that carries them, because the grouping is where the correction below arose.

Federal Reserve, November 2025 (document 5):

  1. “automated, computer-driven trading based on predefined trading strategies”
  2. “self-learning based on historical experience, adaptation based on current market conditions, and analysis of unstructured data, such as text”
  3. “The greater model complexity and the use of additional information by AI currently come at the cost of reduced speed, and thus the suitability of the latest AI models for trading decisions depends on the application.”
  4. “the majority of AI applications in trading today seem to be building upon established practices in machine learning and sophisticated data analysis techniques, rather than representing a significant departure from existing methods”
  5. “efficiency gains, without a fundamental change in the trading process itself, at least for now”
  6. “remained well above its historical median”
  7. “has not necessarily been detrimental to market quality”

Federal Reserve, May 2026 (documents 3 and 4):

  1. “stayed in the upper range of its historical distribution” — overview page
  2. “remained well below its historical average” — overview page

IMF Blog, October 2024 (document 6):

  1. “used in only limited ways by actual investors”
  2. “has risen from 19 percent in 2017 to over 50 percent each year since 2020”
  3. “While a typical actively managed equity ETF turns over its holdings much less than once a year, AI-driven ETFs do so about once a month.”
  4. “since 2017 and the introduction of LLMs, the movement of US equity prices 15 seconds after the release of the Fed minutes seem to be more consistently in the direction of the longer-lasting movement seen after 15 minutes, in contrast to the apparently uncorrelated movements in the pre-LLM period”
  5. “Several AI-driven ETFs saw increased turnover during the March 2020 market turmoil, indicating potential for increased herd-like selling during times of stress.”

IMF Global Financial Stability Report, October 2026 (documents 1 and 2):

  1. “delivered exceptional earnings growth”; “valuation pressure measures such as forward price-to-earnings ratios have eased in several equity markets”
  2. “free cash flows are rapidly declining as hyperscalers try to sustain large AI investment needs”
  3. “reflecting an increase in levered positions in AI-related equities”
  4. “as retail investors have used these funds to gain levered exposures to AI stocks”
  5. “depreciate in value precisely when future revenue streams tied to them are impaired”
  6. “an AI-driven market correction”; “by rising leverage and concentrated exposures”
  7. “growing evidence that AI capabilities can be delivered at lower investment levels”
  8. “by themselves do not necessarily threaten financial stability but could become systemic”; “impairments to financial intermediation and credit provision, with serious negative economic effects”
  9. “the banking system would remain adequately capitalized in a bond yield spike and AI-led correction scenario”; “a significant share of vulnerable institutions would emerge”; “could prove a source of fragility”; “remains relatively low in aggregate”
  10. Figures: 1 per cent to 10 per cent investment-grade issuance share; 10 per cent to about 15 per cent of advanced-economy GDP; 16 per cent to 28 per cent single-stock ETF share; fourteen transactions; $300 billion equity, $650 billion debt, $1.4 trillion future revenue commitments; 0.4, 1.3 and 0.6 per cent scenario.

Verification record

  • Re-tested against live documents on 1 November 2026: every Federal Reserve quotation above, by direct string search of the retrieved pages. All resolved.
  • Re-tested against the IMF Blog text retrieved on 1 November 2026: the patent share, the ETF turnover comparison, the 15-second price-discovery sentence and the March 2020 herd-selling reading. All resolved.
  • Verified against full-text captures: all IMF Global Financial Stability Report material. Both IMF properties returned HTTP 403 to automated retrieval on 1 November 2026, so this material could not be re-fetched. It is verified against captures made before the block and re-tested against those captures. This is a real limit on this file and is stated in the report as well.
  • Corrected during editorial review — attribution: an earlier draft introduced the patent-share, ETF-turnover and 15-second price-discovery figures as “the Fed’s own numbers.” They are the IMF’s October 2024 analysis. The report now attributes them there. The Federal Reserve’s November 2025 box is credited only with the definition, the speed trade-off and the “established practices” assessment. This was the single most material defect found in review.
  • Corrected during editorial review — hedging: the IMF writes that prices seem to move in the direction of the longer-lasting move. An earlier draft dropped the hedge, which turned an observation into an assertion. Restored.
  • Corrected during editorial review — quotation: an earlier draft quoted “the potential for increased herd-like selling during times of stress”; the source reads “potential for increased herd-like selling during times of stress.” The added article is removed.
  • Corrected during editorial review — citation: the May 2026 “upper range of its historical distribution” sentence sits on the report’s overview page, not the asset-valuations page an earlier draft cited alone. Both pages are now cited, which is also why the equity-premium figure and the valuation sentence appear under different URLs.

Deliberate omissions

  • No percentage for algorithmic-trading share of volume. The commonly repeated 60 to 70 per cent figure could not be traced to any primary source. Commercial trackers cite each other; the one authoritative candidate, the SEC’s 2020 Staff Report on Algorithmic Trading, was returned as a PDF whose text could not be extracted in this run. The report states the figure is unsourceable rather than printing it as fact.
  • No percentage for index concentration. Published trackers cluster in the high thirties to low forties for the top-10 share of the S&P 500 but disagree with one another and cite one another. No index-provider or regulator figure was retrieved. The report describes concentration qualitatively.
  • No dot-com valuation-multiple comparison table. No matched dataset was obtained, so the comparison in the report is qualitative and rests on the earnings divergence that the IMF documents.

Stated limits carried into the report

  • The report does not forecast a crash, a date or a magnitude, and says so in its closing section.
  • Two institutional sources (the IMF and the Federal Reserve) are used; there is no independent replication of either body’s analysis and no academic source was introduced.
  • The IMF’s scenario figures are model outputs under stated assumptions, not predictions.
  • Two document sets are age-disclosed in the copy: the IMF report is 18 days old at publication, the Federal Reserve May 2026 data are six months old, and the IMF blog analysis is two years old.

Duplication check

No published article on the site covers AI dependence in equity markets, market concentration, algorithmic trading or the IMF’s financial-stability findings. The nearest published items are a report on Anthropic’s IPO filing and one on European AI investment; neither overlaps this subject matter, and both are linked as related coverage rather than reused.

Provenance of this package

The dossier was researched and drafted with a mechanical verification harness that re-tests every quotation and figure against the retrieved source text. That harness was re-run for this editorial review and returned 44 of 44 assertions held. Its limitation is that it verifies text against captures, not live pages; the live re-testing of the Federal Reserve and IMF Blog material in this file was performed separately, by direct retrieval on 1 November 2026. Nothing in this package has been published, pushed to the production branch or deployed.