Starting Regular Use of AI to Assist in Writing Federal Tax Procedure Blog Entries (10/20/26)

As readers of this blog know, I have on rare occasions consulted an AI tool with respect to some issues I have blogged. I tended to dip my toe in and then stay out for a while. I have decided to start dipping my toe in more frequently and deeper.

I have been inspired now by Robert Wright’s The God Test: Artificial Intelligence and Our Coming Cosmic Reckoning, Amazon here, I have started reading the book and am impressed so far. Fittingly, I asked an AI tool, Microsoft Copilot the following:

Please point me to the best reviews of Robert Wright's The God Test: Artificial Intelligence and Our Coming Cosmic Reckoning and summarize the key points of the reviews.

Readers of this blog entry can read Copilot’s response here.

I have been a Robert Wright fan for a long time—since reading his book titled The Evolution of God , Amazon here, when it came out in 2009. (As an aside, shortly after it came out, I taught a 4-week Sunday School class at South Main Baptist Church, Houston, TX on this book.) Judging from these two book-end titles, one can infer that he thinks a lot about God and, from the content of The God Test, about the continuity of evolution.

Suffice it to say that I now am convinced that I need to use AI more to understand its possibilities and its limitations. So, what better way to start the process than to ask an AI tool to assist me. I used ChatGPT Plus (“ChatGPT”) for this effort. I started with the following prompt:

Tax Court Decides 2 Bullshit SCE Tax Shelter Cases and, in One, Judge Lauber Calls Out Petitioner's Counsel's Theatrics (10/19/26)

The Tax Court decided two syndicated conservation easement (“SCE”) cases today (presented in order of most pages):

• Del Monte, LLC v. Commissioner, T.C. Memo. 2026-107 (67 pages; Docket ## 3411-21 at 381 & 6084-21 at #372, GS here [to come] authored by Judge Urda).
• Watson Lake Investments, LLC v. Commissioner, T.C. Memo. 2026-106 (53 pages; Docket ## 13696-20 at 165 & 24239-22 at 132; GS here [to come], authored by Judge Lauber).

Both of these are standard fare bullshit SCE gross overvaluation cases. I could just let it go at that because readers will already infer correctly that the partnerships presented bullshit valuations on the returns and with so-called expert testimony at trial.

But Judge Lauber’s opinion in Watson Lake deserves a special mention. First, Judge Lauber calls out one of the petitioner’s (Watson Lake’s) reliance on an obviously forged letter. (See Slip Op. 11 & 31-32.). Second and more importantly, Judge Lauber addresses in some detail the petitioners’ counsel’s theatrics (also properly characterized as bullshit in the extreme) to avoid the foregone conclusion that his knowledge of the case should have told him his clients would and should lose.

On counsel’s bullshit arguments, Judge Lauber first handily rejects the constitutional claims. (Slip Op. 43-44.) Nothing worthy of comment.

On counsel’s bullshit theatrics (Slip Op. 44-53), Judge Lauber provides the detail ad nauseum to show many footfaults (a euphemism in this context) by petitioner’s counsel. I copy and paste Judge Lauber’s introduction (Slip Op. 44-45):

ChatGPT and Loper Bright Misread Chevron Footnote 11 to Have Permitted Chevron Deference Over an Opposing Best Interpretation (10/7/26; 10/8/26)

Yesterday, I posted on Loper Bright’s Cryptic Statement of Statutory Stare Decisis for Chevron Precedents – And Giving Chevron a Bad Rap (Procedurally Taxing Blog 10/6/26), here. After I posted, I asked ChatGPT Plus (hereafter just ChatGPT) the following:

This is a blog entry that I wrote today. Please re-write it to make it tighter and cite appropriate authority. [I copied and pasted the blog entry]

Yesterday evening I posted at the bottom of the blog a link to a pdf of the ChatGPT comments and re-write along with some of my comments to ChatGPT’s response. I also offer that pdf link here and the ChatGPT link is here (omitting my comments). I think one of ChatGPT’s critiques of my blog should be elevated to a separate blog because it is so important. Here is the key critique having to do with Chevron footnote 11 (note ChatGPT is speaking in the first person).

I would revise the final argument about Chevron. Chevron itself, especially footnote 11, expressly contemplated upholding an agency interpretation even when it was not the interpretation the court independently would have selected. Your stronger empirical point is that this theoretical feature of Chevron should not be confused with proof that courts actually and frequently rejected what they regarded as the best interpretation in favor of an inferior agency interpretation. Barnett and Walker do not establish that proposition.

Back to JAT speaking for myself in the first person:

1. The key point I [JAT] was making was that Chevron required ambiguity where a court could determine only that competing interpretations were in the zone of ambiguity of the statutory text but could not determine which interpretation is best. So, with that point (which really when carefully considered covers the rest of my analysis), here is the rest:

2. ChatGPT Plus rightly points out that Chevron footnote 11 said: “The court need not conclude that the agency construction was the only one it permissibly could have adopted to uphold the construction, or even the reading the court would have reached if the question initially had arisen in a judicial proceeding.” (467 U.S., at 843 n. 11.) My problem is that, in the Chevron text to which that footnote applies, the Court says and I repeat the text and footnote to contextualize (I bold-face key words)

Loper Bright’s Cryptic Statement of Statutory Stare Decisis for Chevron Precedents – And Giving Chevron a Bad Rap (10/6/26)

I have just reviewed Eli Nachmany, Deference Undisturbed, 101 Notre Dame L. Rev. 985 (2026), on SSRN here. The author focuses on the rather cryptic discussion of statutory stare decisis in the Opinion of the Court (C.J. Roberts) in Loper Bright Enterprises v. Raimondo, 603 U.S. 369 , 412 (2024), here [Preliminary Print]:

By doing so [leaving deference behind in Loper Bright], we do not call into question prior cases that relied on the Chevron framework. The holdings of those cases that specific agency actions are lawful—including the Clean Air Act holding of Chevron itself—are still subject to statutory stare decisis despite our change in interpretive methodology.

The actual holding of Loper Bright was that, going forward, courts must interpret statutes in agency interpretation cases de novo rather than deferring to agency interpretations as Chevron (and, really, Chevron predecessors dating to before the APA) allegedly allowed. Loper Bright's cryptic discussion of statutory stare decisis is dictum rather than a holding and seems to be something textualists often claim to avoid—permitting consequences of the interpretation to affect the interpretation of the text that, under their claimed textual preference looks only to the text, the whole text, and nothing but the text. (As Justice Kagan famously noted, textualists find a way to avoid text when they choose to, using the Major Questions Doctrine, for example, as a “get out of text free” card. West Virginia v. EPA, 597 U.S. 697, 779 (2022) (Kagan, J., dissenting).)

No one can doubt the disruption to a host of precedential statutory interpretations that would occur if Chevron deference could be litigated anew based on Loper Bright. As Nachmany discusses in the article (p. 1010), that cryptic statement was to stem the tide of the disruptive possibility.

Of course, the statement in Loper Bright is cryptic, offering nothing that helps courts in the future decide which precedents are entitled to statutory stare decisis and which are not. Perhaps it may be like pornography: courts will know it when they see it. See Jacobellis v. Ohio, 378 U.S. 184, 197 n. 2  (1964) (J. Stewart, concurring).

I discuss statutory stare decisis in my article. John A. Townsend, Loper Bright Flip Flops on Chevron Deference: A Tax Lawyer's Perspective, 79 Tax Law. 323, 391-397 (2026), SSRN here. In this posting, I offer some discussion that is in more cryptic in my article but inspired now by Nachmany’s article.

The issue that concerns me is to identify the holding of the opinion a court would use for precedent. I do not address the issue of what levels of precedent Loper Bright means its statement to apply (e.g., Supreme Court, Circuit Courts, or trial courts). At a minimum, everyone should recognize that Supreme Court adoption of agency interpretations under Chevron are precedential opinions for all judges to which the cryptic statement in Loper Bright might apply.

One thing that is certain is that the Loper Bright cryptic statement requires that the precedent in question have applied Chevron deference. I illustrate with two examples.

Example 1. The prior opinion noised about Chevron but when read fairly applied the agency interpretation as the best interpretation rather than deferring to it over an opposing best interpretation. This is a Chevron Step One case. Sometimes teasing out from the prior opinions the real basis for the holding can be difficult, but courts having the claimed facility for statutory interpretation should have some skills for interpreting potentially precedential opinions. (In this context the prior opinion is the law requiring interpretation.) As I discuss in my article, many opinions, perhaps a majority, that some view as Chevron cases really are not when analyzed closely because they found the agency interpretation the best. (See discussion in the article at 339-341, particularly at nn. 62-63.) Also, I have previously reported on significant data sets of Courts of Appeals “Chevron” holdings identified by others and found that clear Chevron holdings were rare and some likely approved best agency interpretations even though they noised about Chevron. Chevron Step Two Reasonableness and Agency Best Interpretations in Courts of Appeals (Federal Tax Procedure Blog 2/9/23), here; and Is Chevron on Life Support; Does It Matter? (Federal Tax Procedure Blog 4/2/22; 4/3/22), here. (Focus in each blog on Category 5.) This type of precedent should clearly qualify for statutory stare decisis as to the holdings that (i) the statute is not ambiguous and (ii) the agency interpretation is the best interpretation. These cases are not Chevron cases.

Adventures in AI and Federal Tax Procedure on § 7434 (10/3/26)

Today, I was reviewing a recent decision from the Seventh Circuit: Jacoby v. Idlewild Country Club, No. 25-1608 (7th Cir. Oct. 2, 2026) (nonprecedential disposition), CA7 here and TN here, and added it to a footnote in the § 7434 discussion in my working draft for the 2027 edition (to be published in early August 2027). Since I am trying to develop some skills in AI (even though I have not mastered my own intelligence, if any), I decided to use the § 7434 discussion with revisions to date to test what AI has to offer. I used ChatGPT Plus (a paid version). The results offered may be viewed here. I then took the version as ChatGPT rewrote it, pasted it into WordPerfect (my word processor of choice because I am an old lawyer who has not learned MS tricks), and then refined ChatGPT Plus’s offering. So, the files relevant to this exercise are:

1. The version submitted to ChatGPT Plus from the working draft for 2026. Here. This version has 4,750 words (per WP word count). The red font text indicate revisions in the working draft from the 2026 editions.
2. The version ChatGPT Plus offered (with the proposed footnotes as endnotes). Here. This version has 6,079 words (per WP word count).
3. My changes to the ChatGPT Plus that I have not decided whether to replace in the 2027 edition. Here. This version has 6,757 words (per WP word count).

I think each version has the same information as presented in the original from the 2027 working draft (¶ 1).

I would appreciate those having the interest to give me feedback as to which versions are better and more specifically if the two revised versions (¶ 2 and 3 above) are better in the sense of how they better convey the information in the original version (¶ 1) and are worth the additional words to convey the information.

Tax Court Issues a Precedential Opinion (T.C.) Holding that Form 872-T Does Not Revoke a Proper Form 872 Consent to Extend the Statute of Limitations; Well, Duh! (9/30/26; 10/6/26)

The Tax Court today issued a regular opinion in Fine v. Commissioner, 167 T.C. ___, No. 13 (2026) (T.C. # 24390-21 at #43 and GS here. The issue was, to use tax jargon, whether a taxpayer can use Form 872-T to terminate a Form 872 consent to extend the statute of limitations. OK, to be more precise, Form 872 is a Consent to Extend the Time to Assess Tax; Form 872-T is Notice of Termination of Special Consent to Extend the Time to Assess Tax. (Bold-face supplied by JAT.) On its face the name alone should tip one-off that it may not apply to Form 872, because Form 872 is not nominated a “Special” Consent. Reading Form 872-T, one can quickly determine that it applies to the Form 872-A, nominated Special Consent to Extend the Time to Assess Tax (bold-face supplied by JAT).

I dare say that only the most unaware tax practitioner would have known that a Form 872-T cannot revoke the date certain in a Form 872. But apparently such an unaware tax practitioner was the Fines' lawyer (both with respect to deploying / filing the Form 872-T (5 times) and making the argument in the Tax Court case (see Slip Op. 4 n. 5 identifying the person submitting the form as “individual is also counsel of record for petitioners in this case.”).

So, why would the Tax Court use a regular precedential opinion to state a truism known to all minimally competent tax lawyers? In my mind, this opinion should have been issued as an order and not even elevated to T.C. Memo. status. (But they are Tax Court Judges, and I am not.)

 For readers’ further consideration, I offer the following:

TFRP Cannot Exceed the Employer's Compromised Trust Fund Tax Liability On A Record That Does Not Explain the Compromise (9/28/26; 9/29/26)

In Amodio v. Commissioner, T.C. Memo. 2026-96, T.C. Dkt. 9959-22L at # 50 9/28/26 and GS here [to come], the Court held that, on the facts, the IRS cannot collect more on the Trust Fund Recovery Penalty  ("TFRP") in § 6672 than the employer and the IRS had compromised the underlying liability for. The Court reasoned that the TFRP is derivative of the employer’s liability, so that, absent some special consideration, the TFRP cannot exceed the employer's compromised liability.

On the facts presented in the opinion, that holding seems correct.

However, what is missing is why the employer’s liability was compromised. Generally, the IRS compromises tax liabilities based upon doubt as to liability and doubt as to collectibility. If the compromise had been based on doubt as to liability, Amodio seems right to me. If, however, the compromise had been based upon doubt as to collectibility, Amodio seems wrong to me. 

The problem in Amodio was that nobody explained the basis for the compromise. The IRS just failed to put on the evidence uniquely within its control and thus Amodio is correct on the record as described in the opinion.

Added 9/29/26 9:00am:

I asked ChatGPT Plus to critique the portion of the Amodio opinion dealing with the issue discussed above. For those interested in ChatGPT Plus's critique, see here and here I gave ChatGPT Plus two prompts, the second following through on the first. ChatGPT Plus's responses are lengthy but seem to me to add nuance beyond the cryptic discussion in the opinion.

Article on Using Machine Learning for Predicting Tax Ownership (9/27/26)

I spent more time than I should have reading Young Ran (Christine) Kim and Dmitry Erokhin, Algorithmic Tax Ownership, 51 BYU L. Rev. 1437 (2026), here. Don’t get me wrong, I think the article’s discussion of its premises is quite good. What threw me off was the discussion of the “Algorithmic” part of the article. The authors deploy math and computer modeling to predict tax ownership by digesting a large dataset of cases that have dealt with tax ownership.

Of course, the authors mention prominently the iconic case from my time at DOJ Tax: Frank Lyon Co. v. United States, 435 U.S. 561, 582–83 (1978), GS here. I had an indirect role in Frank Lyon (which I’ll tell in the digression below).

Frank Lyon is the classic Supreme Court screw-up in tax law. For further discussions of the screw-up, see the citations in the article. (p. 1446 n. 22 and 1447 n. 32). One of my favorite articles is Charles I. Kingson, The Confusion over Tax Ownership, 93 TAX NOTES 409, 410 (2001), SSRN here (criticizing Frank Lyon as “nobody knows what it means”). Shortly after leaving DOJ Tax for private practice in 1997, I advised a bank wanting to do for a lead shareholder essentially the same deal as Worthen Bank did for Frank Lyon through his corporation. My advice was essentially that it’d be a crap shoot because nobody knew what Frank Lyon meant, but that we could clone the documents in Frank Lyon to recreate the exactly same structure and to be even more sure the bank should change the bank's name to Worthen Bank and his corporation's name to Frank Lyon Corp. Of course, that was hyperbole, but not much because the multi-factor test of tax ownership that Frank Lyon deployed is less than clear in their application in other cases.

Back to the article, it concludes as follows (p. 1494):

To refine tax ownership analysis, this Article applies machine learning, specifically focusing on repo transactions. It identifies economic benefits and burdens  Factor 1) as the most decisive factor in classifying transactions as either sales or loans, with 94.6% accuracy. Additional factors, while sometimes helpful, do not significantly impact classification outcomes.

Earlier (p. 1484):

Nevertheless, the proposed model undeniably reshapes the current tax ownership analysis by showing the decisive factor of the analysis is who bears the economic benefits and burdens of the underlying asset. The clarity this model brings to tax ownership analysis is applicable in a wide array of contexts including repo transactions, sales and leasebacks, litigation financing, and more.

ChatGPT Plus on Loper Bright's Superficial Claim of Court's Ability to Determine a Single Best Reading in State of Ambiguity (9/24/26; 9/25/26)

In my recent article, John A. Townsend, Loper Bright Flip Flops on Chevron Deference: A Tax Lawyer's Perspective, 79 Tax Law. 323 (2026), on SSRN here, I critiqued the Court’s opinion in Loper Bright Enterprises v. Raimondo, 603 U.S. 369 (2024) (C.J. Roberts), here [Preliminary Print]. One of the points I made was that the Court’s claim that “statutes, no matter how impenetrable, do— in fact, must—have a single, best meaning.” (603 U.S., at 400.) The Court also claimed that courts could and must discern that single best meaning without a default rule to decide the case—such as Chevron and, presumably, some substantive canons that require ambiguity. In Loper Bright’s sweeping claims, no room seemed to be left for statutory ambiguity. However, statutory ambiguity was a common phenomenon before Loper Bright; Loper Bright cannot by fiat eliminate statutory ambiguity.

Added 9/24/26 8:00pm: In my Federal Tax Procedure Book 2026 (Practitioner Ed. p. 91; Student Ed. p. 58) this issue is the first issue I raise about Loper Bright:

1. What does a court do when, after it applies Loper Bright’s exhortations to determine the “best” interpretation, the court still finds statutory ambiguity (interpretive equipoise)? Ambiguity (interpretive equipoise) existed before Loper Bright and exists after Loper Bright. Remember the lesson of lenity–ambiguity (interpretive equipoise) requires a default rule to resolve the case.

I decided to use ChatGPT Plus to further explore the phenomenon of statutory ambiguity. That AI tool gave surprisingly cogent answers and step by step reasoning that highlight the lapse in logic of Loper Bright. I wish I had the responses in preparing the article cited above. I offer below the last two prompts and response (after refining earlier prompts and responses). I have left the bold-face and italics as ChatGPT had them; also, ChatGPT sometimes refers to Loper Bright as Loper and does not italicize Loper or Chevron. I have not changed that. Remember these are the final prompts and responses after earlier prompts drilled down to these final two prompt. Those interested in seeing the entire discussion with ChatGPT Plus may do so here.

I am impressed with how ChatGPT responds in an analytically meaningful way to step through  the issues and reach its conclusions. To readers interested in this stuff, ENJOY!

Two points about the Responses: 

1. The first response below refers to my article above without citing the article or otherwise attributing it to me. See par. 4 below. I do not know if ChatGPT Plus recognized that the article that it cited was written by the person with whom it was "chatting." I infer that it did not recognize me as the author of the article.
2. The first response (par. 5) refers to the Flight Options case which it discussed in earlier responses and which I discuss in Interesting Concurring Opinion on Canons (or Maxims) of Statutory Interpretation (Federal Tax Procedure Blog 5/29/26; 8/21/26), here.

JAT Prompt #1: 

In the foregoing discussion, you used 50-50 as the state of interpretive equipoise. Can interpretive equipoise be encountered in a range--say of 45-50 or even 40-60. (These numbers are in percentages.)

ChatGPT Response to Prompt # 1:

Yes—but I would refine the terminology. Interpretive equipoise need not mean a mathematically exact 50–50 tie. It can sensibly describe a zone of interpretive uncertainty within which a court cannot determine, with sufficient confidence, that one interpretation is genuinely “better” than the other. There is, however, no judicially established numerical range such as 45–55 or 40–60.

That distinction becomes important under Loper Bright.

District Court Sustains IRS Summons at Treaty Partner's Request, Requesting Third-Party Contact Requirement (9/19/26)

In United States v. Yoon (N.D. Cal. No.26-cv-01423-VC  Order dtd 9/17/26 at # 34), CL here and GS here, the Court enforced an IRS summons issued in response to a request by a treaty partner, South Korea. The significant issues addressed are:

1. The Government need only satisfy the four-part Powell test. United States v. Powell, 379 U.S. 48, 57–58 (1964). (Slip Op. 3.)

2 In applying Powell, only the IRS’s good faith is in issue, not the treaty partner’s good faith. Id.

3 Relevance of the summons to a legitimate tax investigation of the treaty partner is not in issue. (Slip Op. 4-5.) South Korea’s request to the IRS “said that all that information was relevant, and it only sought records from the years that match its investigation;” the “request seems sensible on its face.” There is no requirement that the IRS go behind the treaty partner’s statement of relevance.

4. The Court rejected Yoon’s argument that § 7602(c) requires 45-day notice which was not given here. (Slip Op. 5-7.) The notice period, although stated without limitation in the statutory text, is properly read in context not to apply to a summons issued at a treaty partner’s request. The Court concluded (Slip Op. 7) that “there is no ambiguity” as to that reading of the statute.

5. With regard to this interpretation of the statute, the Court acknowledged the IRS’s long-time consistent interpretation to that effect in Regulation § 301.7602-2(c)(3)(C), noting that the interpretation might in a state of ambiguity be useful in statutory interpretation (absent here), this consideration (Slip Op. 7-8):

          For what it’s worth, another consideration works against Yoon here: the IRS’s own longstanding interpretation that section 7602(c) does not apply to foreign tax liabilities. 66 Fed. Reg. 77 (Jan. 2, 2001) (codified at 26 C.F.R. § 301.7602-2(c)(3)(C)). After considering the ordinary meaning of statutory language, a court may look to the interpretation of the agency responsible for implementing the statute. Leon-Vriviesca, 179 F.4th at 1191. Such agency interpretations receive “‘due respect,’ but not binding deference.” Lopez v. Garland, 116 F.4th 1032, 1039 (9th Cir. 2024) (quoting Loper Bright Enterprises v. Raimondo, 603 U.S. 369, 403 (2024)). The IRS’s regulation was issued in 2001, just a few years after the codification of section 7602(c) in 1998, and has remained consistent since then. Therefore, even if the statutory text were ambiguous, the IRS’s interpretation would be “especially useful in determining the statute’s meaning”—and that might cut against Yoon’s legislative purpose argument. Loper Bright, 603 U.S. at 394 (citing Skidmore v. Swift & Co., 323 U.S. 134, 140 (1944)).n5
   n5 The government makes an alternative argument that the general grant of rulemaking authority to the IRS in 26 U.S.C. § 7805(a) empowers the agency to “fill up details” in the statutory scheme; i.e., that section 7805(a) explicitly authorizes the agency to define the phrase “tax liability” in section 7602(c). But the best reading of the statutory text, informed by due respect to the IRS’s regulation, is that section 7602(c) does not apply to foreign tax liabilities. It is therefore not necessary to consider whether section 7805(a) can bear the weight that the government says it does.

JAT Comment: