The Financial Accounting Standards Board's Q1 2026 exposure draft on AI disclosure represents a watershed moment in US accounting standards development.

For the first time in GAAP history, companies are being asked to specifically disclose how artificial intelligence is used in material financial reporting processes, what models are deployed, what data they process, how human oversight is structured, and what uncertainty exists in AI-generated estimates.

The implications span every level of the finance function: CFOs must understand the disclosure framework; controllers must assess which AI uses are "material" and require disclosure; internal auditors must evaluate AI as part of ICFR testing; and audit committees must satisfy themselves that AI governance is adequate.

This guide provides the plain-English framework that all four audiences need.

What FASB's AI Disclosure Exposure Draft Proposes

FASB's exposure draft, formally titled "Proposed Accounting Standards Update: Disclosure of Artificial Intelligence in Material Financial Reporting Processes", was issued for public comment in Q1 2026. The comment period closes in Q2 2026, with a final standard expected in Q3 2026 and an effective date proposed for fiscal years beginning after December 15, 2026.

The exposure draft proposes disclosures in two categories: qualitative and quantitative.

Qualitative disclosures must describe:

) the nature and purpose of AI use in each material financial reporting process
a description of the AI model type (machine learning, large language model, statistical model) and the data inputs it processes
the human oversight and review processes applied to AI-generated outputs before they affect reported amounts
known limitations of the AI system and how those limitations are managed

Quantitative disclosures must include: (1) for AI-generated accounting estimates (loan loss reserves, asset impairments, warranty reserves), a quantification of the sensitivity of the estimate to the AI model's key assumptions; and (2) a comparison of AI-generated estimates to actual outcomes over the prior reporting period where that information is available.

"FASB's AI disclosure proposal is the most significant US GAAP development since ASC 842. Every public company using AI in material accounting processes must begin compliance preparation immediately.", Deloitte Technical Accounting, January 2026

Which AI Uses Trigger Disclosure Requirements

The critical threshold question for controllers and CFOs is: which AI uses are "material" and therefore require disclosure? FASB's exposure draft provides a principles-based definition, AI use is material if its outputs "directly or indirectly affect the recognition, measurement, or disclosure of amounts in the financial statements or the footnotes."

Based on Deloitte and PwC's Q1 2026 client interpretations, the following AI applications almost certainly trigger disclosure requirements under the proposed framework:

Revenue Recognition AI: Any AI model used to determine transaction price allocation, variable consideration estimation, or standalone selling price assessment, common in SaaS, healthcare, and manufacturing companies using AI for contract analysis.
Allowance for Credit Losses (CECL Models): Machine learning models used in CECL calculations for financial institutions and companies with significant trade receivables portfolios, already a significant AI use in financial services.
Goodwill and Asset Impairment Testing: AI models used to forecast cash flows in DCF analyses underlying impairment assessments, or to estimate fair values of intangible assets.
Lease Classification and Modification Analysis: AI tools used to process large lease portfolios under ASC 842, particularly for modification analysis and incremental borrowing rate estimation.
Warranty and Litigation Reserve Estimation: Statistical or ML models used to project expected warranty claims or estimate litigation settlement probability distributions.

AI uses that do not trigger disclosure requirements under the proposed framework include: AI used for internal analytics that don't affect reported amounts, AI writing tools used for draft commentary (where humans control final disclosure language), and AI used for internal audit or risk assessment that doesn't directly feed into financial statement line items.

The SOX 404 and ICFR Implications

The FASB AI disclosure framework intersects significantly with SOX Section 404 internal controls requirements. PCAOB staff guidance issued in 2026 indicates that AI systems used in processes subject to ICFR assessment should be evaluated as IT general controls, with the same testing rigor applied to any other system that materially affects financial reporting.

AI Use CategoryICFR ImplicationTesting Requirement
Revenue recognition modelIT general control over automated processAnnual control testing required
CECL credit loss modelKey control over significant estimateModel validation + change management controls
Impairment DCF modelReview control over accounting estimateManagement review process documentation
ChatGPT for commentary draftingGenerally not ICFR-relevant if human controls final languageDocument human review process
AP invoice matching AIIT general control over transaction processingAccess controls + completeness testing

How to Prepare: A Controller's Action Plan

KPMG's 2026 guidance on the FASB AI disclosure framework recommends a four-phase preparation approach for controllers and CAOs at US public companies.

Phase 1, AI Inventory: Conduct a comprehensive inventory of all AI tools currently used in any process that affects financial statement amounts. Include both formal enterprise AI deployments and informal ChatGPT or spreadsheet-based ML models used by individual team members.
Phase 2, Materiality Assessment: For each AI use identified, assess whether its outputs directly or indirectly affect recognized amounts or required disclosures. Document the materiality determination with technical accounting analysis.
Phase 3, Documentation Build: For each material AI use, build the disclosure documentation required by FASB's proposed framework: model description, data inputs, human oversight process, validation methodology, and known limitations.
Phase 4, Controls Assessment: Working with internal audit and your external auditor, assess each material AI use for ICFR implications and document the relevant control environment, particularly for AI systems used in significant accounting estimates.
Controller reviewing FASB AI disclosure requirements and accounting standards

For finance teams using AI-powered close automation platforms, the documentation requirements under FASB's proposed framework create additional urgency around vendor documentation, including AI model descriptions, data processing disclosures, and validation frameworks that your AI vendor should be able to provide.

The intersection of FASB's AI disclosure requirements and SOX 404 compliance automation is significant, controllers who are already investing in AI-powered controls documentation will find they have built much of the infrastructure the FASB framework requires.

Controller Priority Alert

FASB's proposed effective date of fiscal years beginning after December 15, 2026 means calendar-year public companies need to have AI inventory, materiality assessments, and disclosure documentation ready for fiscal year 2027 first-quarter filings, which means the preparation work must start now.

Private companies are not exempt from preparing. Although FASB's initial framework targets public companies, private companies with lender covenants requiring GAAP financials, companies with PE or VC investors requiring compliance, and companies planning capital markets transactions should build AI disclosure frameworks now.

FASB AI DisclosureAccounting Standards 2026Financial Reporting AIController ComplianceICFR AI

Frequently Asked Questions

Does FASB's AI disclosure rule apply to private companies?
FASB's proposed exposure draft is primarily targeted at public business entities in its initial scope. However, FASB has indicated that private company considerations will be addressed in a subsequent phase. Private companies with significant lenders requiring audited GAAP financials, PE-backed companies, and companies planning IPOs should begin building AI disclosure frameworks aligned with the public company requirements, both as a best practice and to prepare for the likely extension of requirements to private entities.
Does using ChatGPT to draft financial statement footnotes trigger FASB AI disclosure requirements?
Generally no, under the proposed framework, AI used to draft language that is then reviewed and approved by qualified accounting professionals, where the human controls the final disclosure content, does not trigger disclosure requirements. The disclosure threshold is triggered when AI directly or indirectly determines the amounts or measurements reported, not when AI assists in describing those amounts. However, companies should document their human review processes for AI-assisted drafting to demonstrate that AI is not determining the substance of disclosures.
What should we do right now to prepare for FASB's AI disclosure requirements?
The three most important immediate steps are: (1) conduct a comprehensive AI inventory covering all tools used in any process affecting financial statement amounts, including informal tools like ChatGPT, Excel-based ML models, and vendor systems with embedded AI; (2) engage your external auditor now to discuss how they will approach AI in ICFR assessments for your upcoming audit; and (3) work with your AI vendors to obtain the model documentation, data processing disclosures, and validation frameworks that the FASB framework will require you to disclose.

The Bottom Line: FASB's AI Framework Changes the Rules for Every Finance Team

The FASB AI disclosure exposure draft is not a theoretical future concern, it is a practical, near-term compliance imperative for every US public company using AI in material financial reporting processes. The proposed effective date means that companies must have AI inventories, disclosure documentation, and ICFR control frameworks in place within 18 months.

The controllers and CFOs who move earliest on this, conducting AI inventories, building vendor documentation requirements, and engaging external auditors on ICFR implications, will be significantly better positioned than those who wait for final standard issuance. The preparation infrastructure built for FASB compliance also directly supports the IRS AI audit readiness and SEC disclosure obligations discussed throughout this series.

Related Articles

SOX Compliance Automation with AI: How Controllers Are Streamlining Section 404 in 2026

Internal Controls →

AI Hallucination Risk in Finance: CFO Guardrails for Financial Reporting

AI Risk →

IRS AI Audit Enforcement 2026: What CFOs and Tax Leaders Need to Know

Tax Compliance →

AI-Powered Financial Close for Mid-Market US Companies: 2026 Benchmark Report

Close Automation →