Free reference

Regulatory reporting & AI glossary

Plain-English definitions of the major U.S. bank regulatory reports, plus the AI and controls terms that come up when you bring AI into the reporting process. Each report links to its authoritative FFIEC, Federal Reserve, or Treasury source. Written by a current practitioner.

U.S. bank regulatory reports

Call Report (FFIEC 031 / 041 / 051)Report

The Consolidated Reports of Condition and Income that every U.S. bank files quarterly - a detailed balance sheet, income statement, and supporting schedules. FFIEC 031 is for banks with domestic and foreign offices; 041 and 051 are for domestic-only banks (051 is the streamlined version for smaller institutions).

Official source (FFIEC) ↗

FR Y-9CReport

The Consolidated Financial Statements for Holding Companies - quarterly consolidated financials that bank and savings-and-loan holding companies (and U.S. IHCs of foreign banks) file with the Federal Reserve. It mirrors the Call Report but at the holding-company level rather than the bank level.

Official source (FRB) ↗

FR Y-14 (A / Q / M)Report

Capital Assessments and Stress Testing - granular (often loan-, security-, and counterparty-level) data filed by holding companies and IHCs with $100B+ in assets to support CCAR/DFAST stress testing. Three frequencies: annual (with firm projections), quarterly (granular as-of data), and monthly (retail loan-level). It must reconcile to the FR Y-9C - it doesn't replace it.

Official source (FRB) ↗

FFIEC 009 - Country Exposure ReportReport

A quarterly Federal Reserve report (filed via FRB New York) where U.S. banking organizations report their claims on residents of every country - including the U.S. home country - by counterparty sector and maturity, on both an immediate-counterparty and a guarantor (ultimate-risk) basis. A companion FFIEC 009a captures material foreign-country concentrations only.

Official source (FFIEC) ↗

FR 2510 - Institution-to-Aggregate (I-A) Granular DataReport

A quarterly report filed only by U.S. G-SIBs, on the same consolidation basis as the FR Y-9C, capturing a full balance sheet (assets and liabilities) broken out by instrument, currency, remaining maturity, and counterparty country/sector - including the home country. It implements the Financial Stability Board's common data template.

Official source (FRB) ↗

FR 2590 - Single-Counterparty Credit Limits (SCCL)Report

A quarterly report where Category I/II/III firms (and certain FBOs/IHCs) report their top 50 counterparties by net credit exposure, under the single-counterparty credit-limit rules (Regulation YY, 12 CFR Part 252 Subpart H). It measures the firm's exposure to counterparties - it is not a balance sheet, so issuing a security isn't reported, but a liquidity facility to your own conduit is.

Official source (FRB) ↗

FR Y-15 - Banking Organization Systemic Risk ReportReport

A quarterly report where $100B+ organizations (and every U.S. G-SIB) report the six systemic-footprint components - size, interconnectedness, substitutability, complexity, cross-jurisdictional activity, and short-term wholesale funding. These indicators drive the G-SIB capital surcharge. Most cells are bespoke defined measures, not GAAP balance-sheet pulls.

Official source (FRB) ↗

TIC forms - Treasury International CapitalReport

A family of U.S. Treasury reports (filed with FRB New York as fiscal agent) capturing cross-border holdings of, and transactions in, securities and banking claims/liabilities between U.S. and foreign residents. Includes the B-forms (banking claims/liabilities), SLT, SHC/SHCA and SHL/SHLA (securities), Form D (derivatives), and the TFC foreign-currency forms. "Resident" means location, not nationality.

Official source (U.S. Treasury) ↗

FFIEC 101 - Advanced Approaches CapitalReport

A quarterly report where advanced-approaches banks (generally Category I/II) report regulatory capital and advanced-approaches risk-weighted assets - reconciled to the standardized framework. Firms report the more conservative "lower of" advanced vs. standardized ratio (the Collins floor), so it's an overlay on the primary capital schedules (FR Y-9C HC-R / Call RC-R), not a replacement.

Official source (FFIEC) ↗

FR 2052a - Liquidity Monitoring ReportReport

The Complex Institution Liquidity Monitoring Report - granular liquidity data large banks file (the largest daily, others monthly) capturing cash flows by product, counterparty, and maturity. It supports the LCR/NSFR and the Fed's ongoing liquidity monitoring. Maturity uses the earliest contractual exercise date (unlike other reports).

Official source (FRB) ↗
Reporting concepts

Schedule / line itemConcept

A "schedule" is a section of a report covering one topic (loans, deposits, derivatives); a "line item" is a specific reportable value within it, defined by the instructions. A single product can hit multiple schedules, parts, sub-schedules, and memoranda lines.

Reporting instructionsConcept

The official, public guidance (FFIEC, Federal Reserve, OCC, FDIC, SEC) that defines exactly what each line item must include, exclude, and how to classify items. The instructions are the primary authority - the source of truth, and what every AI answer should be checked against.

Edit checks / validationsConcept

Built-in rules a report must pass before submission - internal-consistency and cross-schedule checks that flag likely errors.

Interseries reconciliationConcept

Tying values across different reports or schedules that should agree (e.g., FR 2510 reconciling to FR Y-9C; FFIEC 101 to Call RC-R; FR Y-15 Schedule E to FFIEC 009). A core reporting control - and a natural fit for AI-assisted cross-checking.

Attestation / sign-offConcept

The formal certification by a responsible officer that a report is accurate and complete. AI can prepare and check; a person still signs.

Materiality / reporting thresholdConcept

A defined level below which an item isn't significant enough to report or correct. Many reports also have panel/scope thresholds that decide whether you file at all (e.g., FR 2510 = U.S. G-SIBs; FR Y-14/Y-15 = $100B+).

AI

LLM (Large Language Model)AI

An AI model trained on vast amounts of text that predicts and generates language - the engine behind tools like ChatGPT, Claude, and Microsoft Copilot.

Prompt / prompt engineeringAI

The instruction you give an AI. Prompt engineering is the craft of writing prompts (role, context, constraints) that produce reliable, useful output instead of vague filler.

HallucinationAI

When an AI states something false with full confidence - an invented citation, number, or rule. The single biggest reason every AI answer in reporting work has to be verified against source.

RAG (Retrieval-Augmented Generation)AI

A technique that feeds the AI relevant source documents (e.g., the actual instructions) at query time, so answers are grounded in those documents rather than the model's memory. Reduces hallucination.

AI agentAI

An AI system that takes multi-step actions toward a goal - reading documents, running steps, producing finished work - rather than just answering a single prompt.

Context windowAI

How much text an AI can consider at once. Inputs beyond the limit get truncated, so what you include - and leave out - matters.

Grounding / citationsAI

Tying an AI's answer to specific source text so you can check it. Essential for reg work: an answer you can't trace, you can't defend in an exam.

Fine-tuningAI

Further-training a model on specific data to specialize it. More involved than prompting or RAG - and usually unnecessary for most reporting tasks, which good prompts plus source documents handle.

Controls & governance

Control walkthroughControls

Tracing a transaction or process end-to-end to confirm you understand how it works and where the controls sit.

Design effectivenessControls

Evaluating whether a control, as designed, would prevent or detect the error it targets - before testing whether it actually operates.

Operating effectivenessControls

Testing whether a control actually operated as designed over a period - the "does it really work, day to day" question.

Reconciliation / tie-outControls

Confirming that two independently derived figures agree (e.g., a report total to the general ledger). A foundational reporting control.

Human-in-the-loopControls

A workflow where AI does the work but a person reviews, judges, and approves before anything is final. The default for control-sensitive use.

Model risk management (SR 11-7)Controls

Federal Reserve / OCC supervisory guidance on managing the risk that a model - increasingly including AI - is wrong or misused. Covers development, validation, and governance.

Audit trail / explainabilityControls

A record of what was done and why. For AI: showing the sources, changes, and rationale behind each step so a reviewer or examiner can follow it.

Segregation of dutiesControls

Splitting a task so no one person controls it end-to-end (preparer is not the reviewer), reducing error and fraud risk.

One control-aware idea a week.

Subscribe free for practical, plain-English guidance on using AI in regulatory reporting - and the guardrails that keep it safe.

Subscribe free