AI for FBAR Compliance: FinCEN 114 + §10.22 Workflow

For solo EAs and small-firm CPAs who already do FBARs every year. The question isn’t whether AI helps. The question is at which workflow stage, with what consent posture, and how to paper §10.22 onto the output.

The 30-second answer #

No AI vendor as of May 2026 ships a purpose-built FinCEN 114 module. The leverage points are five workflow stages — account discovery, statement aggregation, currency conversion, signature-authority detection, and the §10.22 review trail. Three are genuine time compressors; one is AI-assisted but discipline-dependent; one creates the audit-defense artifact. The §7216 architecture governs which vendors a practitioner can legally route client documents through. Post-Bittner, the non-willful penalty is $16,536 per annual form (2025-adjusted); willful exposure is per-account at the greater of $165,353 or 50% of the balance.

Why an FBAR-AI article is different from a generic AI article #

Most “AI for tax” content addresses research and drafting. FBAR is different. The painful parts aren’t research — they’re data extraction from format-diverse foreign bank statements, currency conversion with a non-obvious official rate, and signature-authority discovery buried in employment contracts and POAs that no organizer asks about. Three of those four time sinks compress under AI. The fourth — the §10.22 review trail — is created by AI, in the sense that AI-generated workpapers become the audit-defense artifact.

The vendor landscape reflects the gap. As of May 2026:

Vendor	Relevant capability	FBAR-specific module?	§7216 posture for doc upload
Black Ore	Autonomous 1040 pipeline; ingests “bank records in any format”	No — output routes to Form 1040, not FinCEN 114	SOC 2 Type II; confirm DPA terms
CoCounsel Tax	Research over Checkpoint content; FBAR-adjacent reference material	No workflow integration	Thomson Reuters enterprise; DPA typical
Blue J	Research over IBFD + IRC + Tax Court	No — FBAR sits partly outside its content universe	SOC 2 Type II annual
TaxGPT	General-purpose research assistant	No explicit FBAR coverage claim	Strongest public posture: SOC 2 Type II, no-training, DPA available

Practitioners build the FBAR workflow by stitching general-purpose AI capabilities to specific stages, with the §7216 architecture and the §10.22 documentation discipline doing the load-bearing work.

“FBAR compliance is too often viewed as a straightforward exercise. Real-world complexities, however, challenge even experienced practitioners.” — Virginia La Torre Jeker, J.D., us-tax.org

The article doesn’t automate a form. It systematizes the parts of an underdiagnosed compliance regime that the senior practitioner already knows are traps.

The compliance floor in 90 seconds #

Statutory authority sits at 31 USC §5314. FinCEN — a bureau of Treasury under Treasury Order 180-01 — administers the regime through 31 CFR Part 1010. The $10,000 aggregate threshold appears at §1010.306(c) — aggregate across all foreign financial accounts, at any point during the calendar year. Three accounts of $4,000 each = $12,000 aggregate = FBAR required for all three accounts, not just the ones individually over $10,000.

“Financial account” under §1010.350(c) is broad — bank, securities, mutual funds, insurance policies with cash surrender value, foreign annuities, and similar pooled vehicles. “Signature or other authority” under §1010.350(f)(1) is broader than physical signing — it’s the authority to “control the disposition” by “direct communication,” which includes phone, email, or electronic portal. Joint accounts get each US owner filing separately for the full amount, not 50%; married spouses may elect to file jointly via Form 114a authorization.

Filing is e-only via the BSA E-Filing System. Deadline is April 15 with an automatic extension to October 15 — no form required. The CFR text at §1010.306 still references “June 30” — a pre-2016 artifact superseded by Public Law 114-41 and IRM 4.26.16.3.1. Records under §1010.420 retain for five years from the filing date, with tolling during any criminal proceeding tied to false federal tax filings.

Five workflow stages where AI compresses time #

Divide FBAR preparation into five stages. Identify where AI compresses. Name the vendor or workflow that does the work. Flag the §7216 boundary at each stage. The stages aren’t equally AI-amenable — 1, 2, and 4 are the genuine compressors; 3 is AI-assisted but discipline-dependent; 5 is where AI generates the audit-defense artifact.

Stage	Time compression	§7216 boundary	Output artifact
1 — Account discovery	High — buried across 5-10 documents	Required if client docs uploaded	Verified account list
2 — Statement aggregation	High — format-diverse foreign bank PDFs	Required — foreign bank statements are TRI	Monthly-balance workpaper
3 — Currency conversion	Medium — discipline-dependent	Lower if Treasury table is the only sensitive input	USD-converted balance table
4 — Signature-authority detection	High — buried in 30-60pp legal docs	Required if employment/POA/trust docs uploaded	Extracted-clause memo
5 — §10.22 review trail	The trail itself is the AI deliverable	N/A — internal practitioner workpaper	Audit-defense workpaper

Stage 1 — Account discovery

The starting problem: the FBAR-required account list does not arrive in the client organizer. It’s scattered across the engagement letter, prior-year Schedule B, prior-year FBAR, 1099-INT statements from US institutions (which sometimes flag foreign-source income), employment contracts (signature authority over employer-foreign-subsidiary accounts), financial POAs (signature authority over a parent’s foreign account), trust agreements, and the client’s casual mention of “the account my mother left me in Mumbai.”

AI compresses this materially. The workflow: with a §7216-defensible vendor — TaxGPT or Filed are the cleanest publicly documented cases, with anonymization or no-training contracts — upload the prior-year FBAR, the engagement letter, the prior-year Schedule B, and any organizer responses. Prompt for extraction of foreign financial accounts, account numbers, institution names, countries, and any references to signature authority or power of attorney over accounts not personally owned. The output is a draft account list the practitioner verifies against the client interview.

The AICPA’s 2025 FBAR engagement letter contains a structured signature-authority checklist that doubles as an extraction template. Feeding this template to the model alongside client documents narrows the discovery problem to a defined schema.

Stage 2 — Statement aggregation

The format problem: 4-12 statements from foreign banks across as many jurisdictions. European banks variously report in monthly grid, narrative, or transaction-only format. Asian banks may produce statements in two languages with separate balance figures. Manual re-keying of 12 months × 6 accounts is a half-day’s work on a return that earns the practitioner perhaps $750-1,500.

Document-ingestion AI compresses this stage. Filed and Black Ore accept “bank records in any format” — the plumbing handles foreign statements even though neither routes to FinCEN 114. The practitioner extracts structured monthly balances into a workpaper, then populates FinCEN 114 manually.

For practitioners without an enterprise pipeline, frontier-model LLMs (Claude 3.5 Sonnet, GPT-4o, Gemini 1.5 Pro) handle table extraction well — generally, when the §7216 architecture is correct. The prompt has to be explicit about FBAR’s specific metric: highest balance during the year, not year-end. Foreign banks routinely show only month-end balances; the practitioner needs to flag missing daily-peak data and request supplementary statements where the maximum value can’t be derived from the monthly grid.

Stage 3 — Year-end currency conversion

FinCEN 114 requires conversion using the Treasury Reporting Rates of Exchange — not the IRS average rate, not Google’s spot rate, not the foreign bank’s quoted USD figure. The official source is the Treasury Fiscal Service dataset, published quarterly. For FBAR, the December 31 rate applies; the 2025 year-end rates were published January 15, 2026.

AI is genuinely useful here but only with discipline. The model must be given the Treasury rates table as context — pasted into the conversation or uploaded as PDF. It must not be allowed to source its own rates. With the table in context, the conversion pass produces a workpaper-ready table showing currency, highest balance native, Treasury rate applied, USD equivalent, and citation back to the Treasury table. Without the table in context, the model will fabricate or substitute Google’s rate. This is the most discipline-sensitive AI step in the workflow.

Stage 4 — Signature-authority detection

This is the most under-discussed stage, and the one where AI offers the largest compression. The §1010.350(f) “signature authority” rule extends FBAR to US persons who do not own a foreign account but can “control the disposition” of it. Common traps: the CFO who authorizes wires from the foreign subsidiary’s operating account; the US child holding power of attorney over a parent’s foreign bank account; the trustee with discretion over foreign-trust-held accounts; the investment manager with discretionary trading authority.

These triggers sit buried in 30-60 pages of legal documents — employment contracts, POAs, trust agreements, marital property settlements, corporate board resolutions — that no tax organizer asks about. Manual review at this depth is rare. Most signature-authority FBAR exposure surfaces only after IRS notice.

AI compresses this stage by reducing the manual-read burden. Upload the document and prompt the model to identify all language granting authority to control, direct, or instruct transactions in any financial account, with attention to foreign financial institutions named or jurisdictions referenced. The model returns extracted clauses; the practitioner verifies and documents.

The institutional-employee exception under §1010.350(f)(2) — for officers of banks examined by OCC, FDIC, Federal Reserve, OTS, or NCUA, and for SEC- or CFTC-registered institutions, with no financial interest in the account — survives this analysis only if the practitioner applies the “no financial interest” qualifier deliberately. LLMs typically state the general rule and miss the qualifier in edge cases. Mitigation: prompt the model explicitly for both the exception and the qualifier, then verify against the client’s compensation structure.

FinCEN Notice FIN-2025-NTC3 (December 8, 2025) extends the filing deferral for certain financial-industry professionals through April 15, 2026 / April 15, 2027. The deferral is narrow — confirm the client falls inside the §1010.350(f)(2)(i)–(v) class before treating it as covering the case. The deferral has been re-extended every year since 2011; each year’s notice has to be re-verified for the specific client situation.

Stage 5 — The §10.22 review trail

“The tax practitioner employing tools remains responsible for the completed work product” and practitioners should “take reasonable steps to determine that the tools used are appropriate for their intended purpose.” — Edward R. Jenkins, CPA, CGMA, The Tax Adviser, February 2024

The IRS Office of Professional Responsibility made the FBAR-specific Circular 230 application explicit. OPR Alert 2025-10 (July 2025) establishes that practitioners preparing 1040, 1065, or 1120 returns must inquire of clients with “sufficient detail to ascertain the information necessary to enter correct responses” to the Schedule B (and equivalent) foreign-account questions; must promptly inform clients of FBAR noncompliance discovered in the course of the engagement; and have an affirmative obligation to advise clients of the FBAR requirement and its consequences even when the practitioner does not personally prepare the FBAR. The duty extends to identifying noncompliance, error, or omission in past filings. Jenkins’s framing applies the same standard to AI tools: the practitioner remains the signer, regardless of which “person(s) performing the work” — junior staff or model — generated the inputs.

That generates a documentation obligation that AI actually helps fulfill. For each AI-assisted step — discovery, aggregation, conversion, signature-authority scan — the workpaper should capture the prompt used, the model and version, the date, the AI output verbatim or summarized, and the practitioner’s verification or override. This is the audit-defense artifact. It’s also the artifact that demonstrates “reasonable steps to determine that the tools used are appropriate” — a defensible §10.22 trail that ad-hoc paper notes do not produce. The companion analysis at Circular 230 and AI walks the broader §10.22 framing.

A workable workpaper schema — one row per AI-assisted step, papered to the engagement file:

Stage	Tool	Model + version	Date	Prompt summary	AI output summary	Practitioner action
3 — Currency conversion	Claude 3.5 Sonnet (DPA in place)	claude-3-5-sonnet-20241022	2026-04-08	”Using only the attached Treasury Reporting Rates of Exchange as of 12/31/2025, convert each highest-balance figure to USD.”	4-row USD table with Treasury rate cited per currency	Verified each conversion against attached Treasury PDF; one rate adjusted (CHF — used 12/31 rate not Q3 rate)

Same schema applies to Stages 1, 2, and 4. The discipline is filling a row each time AI touches the engagement — not the format.

“Document good-faith Streamlined eligibility analysis. Maintain written audit trail of FBAR reviews and warnings.” — Anthony N. Verni, J.D., CPA, MBA, VERNItaxlaw.com

Verni’s recommendation predates the AI-tool emphasis but applies directly. The audit trail is what protects the practitioner; the AI is what makes the trail consistent across engagements.

Where AI gets FBAR wrong #

Four failure modes survive direct testing of frontier-model LLMs against FBAR questions. Practitioners building AI into the workflow should pre-bake the mitigation into prompts and review steps.

Currency conversion source. LLMs default to Google’s spot rate or the IRS annual average rate. The Treasury Reporting Rates of Exchange are the FBAR-correct source. Mitigation: always provide the Treasury rates table as context. Direct-question testing (Claude 3.5 Sonnet, May 2026) returned the correct methodology when asked plainly — but in-document extraction with the table absent reverts to fabricated or Google-sourced rates.

Highest-balance vs year-end balance. FBAR requires the highest balance at any point during the calendar year, not the December 31 closing balance. An account holding $150,000 in July that closes at $8,000 in December is reportable, and the $150,000 figure is what gets reported. Most foreign statements show only month-end balances; daily-peak data may need to be requested separately. The mitigation is explicit prompt language: “highest balance at any point during the year, not year-end.”

Joint account treatment. Each US person joint owner files separately, reporting 100% of the account value. There is no 50% split. LLMs commonly default to one-filing-covers-both for spouses (which is available only via the Form 114a election) or 50%-each for non-spouses (which is wrong in all cases). HNW practices with sibling-held accounts, parent-child joint accounts, or business-partner foreign accounts hit this regularly.

Crypto stale-data risk. LLMs trained on 2022-2023 data absorbed news of the FinCEN crypto NPRM and may state that foreign virtual-currency accounts are FBAR-reportable. They are not, under current law. FinCEN Notice 2020-2 controls; the NPRM has not finalized as of May 2026. Mitigation: feed the Notice text as context when asking crypto-FBAR questions. Don’t rely on training data on this point.

Bittner and what the penalty math means for AI workflow #

United States v. Bittner, 598 U.S. 85 (Feb 28, 2023), settled the per-form / per-account question for non-willful FBAR violations: penalties accrue per annual report, not per account. The vote was 5-4 — Gorsuch (author) plus Roberts, Alito, Kavanaugh (except Part II-C), and Jackson, against Barrett (author) joined by Thomas, Sotomayor, and Kagan. The cross-ideological lineup matters: the holding rests on statutory text interpretation, not policy preference.

The 2025 inflation-adjusted figures sit at the operative numbers: $16,536 maximum per annual FBAR form for non-willful, and the greater of $165,353 or 50% of the account balance per account for willful — both effective January 17, 2025 per the FinCEN inflation-adjustment final rule. The inflation adjustments apply to penalties assessed after the effective date, regardless of when the underlying violation occurred. That creates retroactive exposure for pre-2025 violations still in the assessment pipeline.

The willful / non-willful determination is now the central financial-risk variable. Post-Bittner, the IRS has visibly shifted enforcement effort toward establishing willfulness — using proxy accounts, missing US indicia (passport, W-9), and Schedule B false answers as willfulness indicators.

“Since the US Supreme Court’s decision in the Bittner case, FBAR penalties are on the IRS radar,” with the IRS “increasingly classifying noncompliance as willful to maximize penalties.” — Virginia La Torre Jeker, J.D., us-tax.org

United States v. Sagoo (N.D. Tex. October 2025) added a procedural wrinkle: applying Jarkesy doctrine, the court held that the IRS cannot assess willful FBAR penalties without a jury trial. Strategic implications cut both ways — the IRS may push more cases into Streamlined (jury trial unavailable in administrative resolution) or selectively prosecute only cases with clean paper trails. Either response sharpens the value of practitioner documentation.

The Credit Suisse Services AG guilty plea (May 5, 2025) — the first IRS FBAR criminal investigation to produce a bank-level plea, with $510M+ in sanctions across 475+ accounts — closed the era of institutional impunity for foreign banks. The downstream effect for practitioners: foreign-bank cooperation with US enforcement is now near-universal, and information that historically didn’t surface (account-holder identity confirmation, transaction records, US-indicia analysis) is increasingly available to IRS examiners.

“FBAR violations are no longer viewed as ‘civil, minor mistakes.’ They now fall squarely under criminal FBAR enforcement, with real prison time and millions in penalties.” — Anthony N. Verni, J.D., CPA, MBA

Crypto FBAR — the unresolved NPRM and practitioner consensus #

FinCEN Notice 2020-2 (December 31, 2020) confirmed that virtual currency accounts at foreign exchanges are not currently reportable on FBAR and announced FinCEN’s intention to propose a rule that would add them. The intended NPRM has been in regulatory limbo for over five years. As of May 2026, no final rule has been published.

Three practitioner positions exist on this unresolved point:

1. Technically correct, more permissive: Notice 2020-2 controls. Pure crypto foreign accounts are not reportable. Document the Notice 2020-2 reliance position in the file.
2. Conservative: File for foreign crypto accounts above the threshold. The cost of filing is low; the cost of post-finalization retroactive exposure is higher.
3. Hybrid (common in HNW practice): File for mixed crypto+fiat accounts where fiat exceeds the threshold (with full account value reported, including the crypto portion). For pure crypto accounts, document the Notice 2020-2 reliance and advise the client of the pending rule.

“Crypto held directly or on foreign non-custodial platforms is arguably not currently reportable on FBAR.” However, “tax professionals typically advise a conservative approach, recommending reporting foreign crypto accounts despite current ambiguity.” — Virginia La Torre Jeker, J.D., us-tax.org, December 2025

The AI failure mode here is direct: LLMs trained on 2022-2023 data may state crypto is currently reportable. The mitigation is to feed the Notice 2020-2 text as context for any crypto-FBAR question. Training data alone is not reliable on this point; the regulatory state hasn’t matched the model’s training corpus since the NPRM dropped.

The Streamlined edge case — SDOP, SFOP, and the Gyetvay warning #

Streamlined Filing Compliance Procedures sit alongside FBAR as the corrective-compliance path when prior-year filings were missed for non-willful reasons. Two programs, structurally different.

SDOP — Streamlined Domestic Offshore Procedures. For US persons residing in the US (substantial-presence under §7701(b)). Submit Form 14654 with a non-willfulness narrative. Penalty is 5% of the highest aggregate balance across the 6-year FBAR period. Three most recent tax years are amended; six most recent FBAR years are submitted.

SFOP — Streamlined Foreign Offshore Procedures. For US persons who were not US residents during at least one of the three most recent tax years (the 330-day physical-presence test for citizens and LPRs). Submit Form 14653. Penalty is zero.

The residence test is load-bearing. The Gyetvay prosecution (former CFO of Novatek, 86 months in prison for hiding $93M in offshore Swiss accounts and falsifying a Streamlined Compliance filing) is the precedent that defines the cost of misrepresenting eligibility. Streamlined certification is signed under penalty of perjury; a false non-willfulness narrative is itself a felony.

AI helps with Streamlined by structuring the non-willfulness narrative — pulling facts from client correspondence, prior tax returns, and engagement files into a chronological account. AI does not help with the eligibility analysis itself. That’s the practitioner’s professional judgment, and any AI that produces a confident “yes, you qualify” without examining client facts in depth should be treated as a hallucination risk. The audit trail under §10.22 carries the burden of demonstrating the practitioner’s reasoning if the IRS challenges the certification.

The §7216 footgun in the FBAR workflow #

Foreign bank statements are the most sensitive documents in a tax file. Pasting one into a public LLM endpoint is a §7216 disclosure with no defensible exception under Treas. Reg. §301.7216-2. Criminal exposure runs up to $1,000 per offense plus possible imprisonment; civil exposure under §6713 is $250 per violation up to $10,000 annually.

The companion analysis at §7216 and AI consent for client data walks the framework in full. The FBAR-specific application: research questions about FBAR rules with no client data can use any LLM — that’s legal research. The §7216 line is crossed when a specific client’s documents or facts are uploaded or quoted to a vendor without a signed consent or a §7216-defensible vendor architecture.

Vendor matrix for FBAR document work, as of May 2026 (verify before relying on any single row):

Vendor	SOC 2 Type II	No-training policy (public)	DPA available	Verdict for FBAR doc upload
TaxGPT	Yes (certified)	Yes — explicit	Yes (/legal/dpa)	Defensible — strongest public posture
Filed	Yes (compliant)	Yes — anonymization before processing	Not confirmed public	Defensible — Azure infrastructure
Black Ore	Yes (certified)	Not stated explicitly	Not confirmed public	Defensible for prep work; confirm DPA terms first
CPA Pilot	Building toward (not yet certified)	Not stated	Not confirmed	Not currently certifiable
Hive Tax	Not found in public materials	Not stated	Not confirmed	Insufficient public data
Public ChatGPT / Claude.ai / Gemini	N/A (consumer)	No (consumer models train unless opted out via enterprise)	Enterprise plans only	Not defensible on consumer endpoints

The §7216 consent generator produces the consent template a practitioner needs before any vendor sees client data.

Practitioner FAQ #

Does AI eliminate the practitioner’s FBAR liability?

No. OPR Alert 2025-10 (July 2025) reaffirmed that the FBAR-related Circular 230 diligence and inquiry duties sit on the practitioner, not the tool. Jenkins’s framing in The Tax Adviser applies the same standard to AI: the practitioner is responsible for engaging, supervising, training, and evaluating any “person(s) performing the work,” junior staff or model alike. The signature on the return is still the practitioner’s, made under penalties of perjury.

Can I use ChatGPT to research FBAR rules without §7216 exposure?

Yes for general research with no client-specific data. The §7216 line is crossed when client documents or facts are uploaded. “What is the FBAR threshold?” is fine. “Does my client Pat Q. Sample need to file FBAR for these three accounts in Zurich?” is a §7216 disclosure if it’s sent to a vendor without a DPA.

What’s the §7216-safest vendor for FBAR document work right now?

TaxGPT has the strongest publicly documented compliance posture as of May 2026 — SOC 2 Type II certified, explicit no-training-on-customer-data policy, DPA available, US data residency. Filed and Black Ore are plausibly defensible, with DPA confirmation recommended before any document upload. Generic public-endpoint LLMs are not defensible without an enterprise plan.

My client has a foreign Coinbase account holding only crypto. Reportable on FBAR?

Not under current law. FinCEN Notice 2020-2 controls; pure crypto foreign accounts are not FBAR-reportable. The intended NPRM that would have changed this has not finalized as of May 2026. Document the Notice 2020-2 reliance position in the file. If the account holds any fiat at any point above the threshold, the fiat triggers reportability and the full account value (including crypto) gets reported.

Highest-balance during the year vs year-end balance — which goes on FinCEN 114?

Highest balance at any point during the calendar year. This is the FBAR-correct metric and the single most common substantive error. Foreign statements often show only month-end balances; daily-peak data may need to be requested. Practitioners using AI to extract balances should explicit-prompt for the highest-during-year metric and flag missing daily data.

My client and their sibling have a joint account in Italy. Who files?

Each US-person joint owner files separately for 100% of the account value. The 50% split is wrong. Married spouses may elect to file a single FBAR via Form 114a authorization. Non-spouse joint owners — siblings, business partners, parent-child — each file separately.

Is the Bittner ruling stable, or could it be overturned?

Stable. The 5-4 cross-ideological majority (Gorsuch + Roberts, Alito, Kavanaugh, Jackson) rests on statutory text interpretation. The holding addresses non-willful penalties only — willful penalties remain per-account at $165,353 (2025-adjusted) or 50% of the balance, whichever is greater. The willful / non-willful determination is now the central variable.

Does the §10.22 review trail need to follow a specific format?

No. It needs to be discoverable, dated, and to show practitioner verification of AI output. A workpaper recording the prompt, AI response, date, model version, and practitioner action satisfies the standard. The format is the practitioner’s choice; the discipline is the documentation cadence.

Sources and citations #

All URLs verified live as of 2026-05-16. The article was assembled from approximately 34 primary and secondary sources across federal statute, FinCEN regulations, IRS guidance, OPR alerts, court opinions, practitioner publications, AICPA materials, vendor documentation, and Treasury data.

Federal statute and regulations

• 31 USC §5314 — Bank Secrecy Act FBAR authority
• 31 CFR §1010.350 — Reports of foreign financial accounts
• 31 CFR §1010.306 — Filing of reports
• 31 CFR §1010.420 — Records to be made and retained
• Treas. Reg. §301.7216-2 — Permissible disclosures
• Treasury Order 180-01 — FinCEN delegation

IRS guidance

• IRS FBAR overview — reviewed April 9, 2026
• IRM 4.26.16 — Report of Foreign Bank and Financial Accounts, updated August 26, 2025
• Form 8938 vs FBAR comparison — reviewed September 18, 2025
• Streamlined Filing Compliance Procedures — reviewed February 19, 2026

FinCEN guidance

• BSA E-Filing System
• FinCEN Notice 2020-2 — crypto reporting position
• FinCEN inflation-adjustment final rule, FR Doc. 2025-01374 — effective January 17, 2025
• FinCEN Notice FIN-2025-NTC3 — signature authority extension

Office of Professional Responsibility

• OPR Alert 2023-12 — Practitioner Diligence Obligations and FBAR
• OPR Alert 2025-10 — Tax Practice Obligations and FBAR — July 2025

Case law

• United States v. Bittner, 598 U.S. 85 (2023) — per-form non-willful penalty
• Schwarzbaum v. United States, No. 22-14058 (11th Cir. Aug 30, 2024) — covered at The Tax Adviser
• Jenner v. Commissioner, 163 T.C. No. 7 (2024) — covered at The Tax Adviser

Practitioner publications

• Edward R. Jenkins, “Tax Ethics and Use of Generative AI Systems,” The Tax Adviser, February 2024
• Lauren M. Azebu, “Offshore and Out of Mind,” The Tax Adviser, April 2024
• Virginia La Torre Jeker, FBAR webinar announcement, us-tax.org, June 2025
• Virginia La Torre Jeker, The Sagoo Case, us-tax.org, October 2025
• Virginia La Torre Jeker, Offshore Crypto, us-tax.org, December 2025
• Anthony N. Verni, Recent Developments in Criminal FBAR Enforcement
• Russ Fox, FBAR Non-Willful Penalty Is Per Report, Taxable Talk, Feb 28, 2023
• Ed Zollars, FBAR Compliance and the CPA’s Due Diligence Under Circular 230, July 2, 2025

AICPA materials

• 2025 FBAR Engagement Letter

Vendor documentation

• TaxGPT Security
• Filed AI compliance article
• Black Ore Tax Autopilot launch
• CoCounsel Tax
• Blue J IBFD partnership

Treasury data

• Treasury Reporting Rates of Exchange — official FBAR conversion source

Notice an outdated citation or broken link? Email corrections@aitaxpractitioner.com — every source is reviewed at minimum quarterly and updated when underlying authority changes.

Related reading #

• §7216 and AI consent for client data — the full §7216 framework that underlies the FBAR vendor-selection question
• Circular 230 and AI: §10.22 due-diligence on AI output — the broader §10.22 framing this article applies to FBAR
• §7216 consent generator tool — generates the consent template referenced throughout
• Real Estate Professional Status — the other major HNW-practitioner compliance topic this site covers