I ran SECURE 2.0 through ChatGPT, and here's what it produced

Conversations about Artificial Intelligence (AI) seem to be everywhere right now. You hear stories of “amazing outcomes” following the slightest effort. For example, I took my Aldrich headshot, ran it through Lensa, and received about 100 pieces of Neil-headshot-centric art in minutes; some of my favorites can be seen below- more are available upon request.

The undisputed leader of the current AI conversation is OpenAI’s ChatGPT—a truly amazing technology with virtually unlimited application. One colleague recently shared a story about a school board meeting that he attended on the topic of fund raising; a member presented “6 keys to successful fundraising” and the group unanimously agreed that the ideas were first-rate; there was genuine excitement and a powerful call to action- the member then shared that the ideas had come from a simple query into ChatGPT.

There are so many other, equally fascinating stories I could share- and I’m sure you’ve heard quite a few yourself- the ability to get so much out of doing so little, right or wrong, is very intriguing and introduces a whole new world of possibilities (and efficiency gains) for those interested in embracing the technology. For me, an obvious test case was the 400-page package that made up “SECURE 2.0.” Since I would be studying this at length anyway, why not see how close to the mark AI could get in seconds compared to my hours and hours of invested time?

I have my own thoughts about the output, but you be the judge. That said, this should be considered more of an interest piece and not a definitive “Summary Guide to Secure 2.0.”

The original text of SECURE 2.0 can be found here, starting on page 817. Let’s start with Section 101. Here is a very, very small snippet of how that section (which has to do with Automatic Enrollment) appears in its original format:

SEC. 101. EXPANDING AUTOMATIC ENROLLMENT IN RETIREMENT PLANS.

1. IN GENERAL.—Subpart B of part I of subchapter D of chapter 1 is amended by inserting after section 414 the following new section:

“SEC. 414A. REQUIREMENTS RELATED TO AUTOMATIC ENROLLMENT.

“(a) IN GENERAL.—Except as otherwise provided in this section—

“(1) an arrangement shall not be treated as a qualified cash or deferred arrangement described in section 401(k) unless such arrangement meets the automatic enrollment requirements of subsection (b), and

“(2) an annuity contract otherwise described in section 403(b) which is purchased under a salary reduction agreement shall not be treated as described in such section unless such agreement meets the automatic enrollment requirements of subsection (b).

With no additional commands, I took all of Section 101 in its original form and entered it into ChatGPT. This is what ChatGPT returned:

ChatGPT: This section of the proposed legislation amends subpart B of part I of subchapter D of chapter 1 of the US tax code to include a new section (414A) that requires certain retirement plans to have automatic enrollment features. The plans that are affected are those that are described in sections 401(k) and 403(b) of the tax code. The automatic enrollment requirements state that the arrangement must be an eligible automatic contribution arrangement (as defined in section 414(w)(3)), allow employees to make permissible withdrawals, have a minimum contribution percentage of between 3% and 10% (increasing by 1 percentage point each year up to 15%), and be invested in accordance with the requirements of a specific regulation. Exceptions are made for certain types of plans, such as simple plans, plans established before the enactment of this section, governmental and church plans, and new and small businesses.

Next, I used the commonly cited “Summarize for a 2^nd grader” command in ChatGPT and applied that to the Section 101 text, which generated the following:

ChatGPT: This is talking about how companies should make it easier for their workers to save money for when they retire. It says that the company should automatically take a little bit of the worker’s pay and save it for them, unless the worker says they don’t want them to. And as time goes on, the company should save a little bit more. This way, the worker will have more money saved up when they stop working.

Well, that’s not so hard to understand!

At the risk of having this piece become too lengthy, what follows is an alternative look at some of the other provisions, effective in 2023, that I reference in my more in-depth piece about SECURE 2.0.

Below are outputs from ChatGPT on Sections 107 and 302 on Required Minimum Distributions, Section 604 on Roth Employer Contributions, Section 320 on Participant Notices, and Section 312 on Self-Certifying Hardships.

Required Minimum Distributions (RMDs) Age – Section 107

ChatGPT: This section of the law increases the age at which individuals are required to begin taking mandatory distributions from their retirement accounts. The current age is 72, but this law increases it to 73 for those who attain age 72 after December 31, 2022 and before January 1, 2033, and to 75 for those who attain age 74 after December 32, 2032. The changes apply to distributions required to be made after December 31, 2022.

Required Minimum Distributions (RMDs) Penalty–Section 302

ChatGPT: This section of the Act amends the tax code to reduce the excise tax on certain accumulations in qualified retirement plans from 50 percent to 25 percent. It also creates a “correction window” during which a taxpayer who receives a distribution of the amount that resulted in imposition of the tax, and submits a return reflecting the tax, will have the first sentence of the subsection (a) applied by substituting “10 percent” for “25 percent”. The amendments made by this section will apply to taxable years beginning after the date of the Act’s enactment.

Roth Employer Contributions–Section 604

ChatGPT: This section of the bill amends the Internal Revenue Code to allow for the optional treatment of employer matching or nonelective contributions to a qualified Roth contribution program as Roth contributions. These contributions will be treated as matching contributions for purposes of the chapter, but will not be excludable from gross income. It also defines a “matching contribution” as including contributions to an eligible deferred compensation plan by an eligible employer on behalf of an employee and on account of their elective deferral under the plan, as long as it is nonforfeitable at the time received. The amendment will apply to contributions made after the date of the Act’s enactment.

Participant Notices–Section 320 (pretty interesting if you compare to the original text)

ChatGPT: This section of the Employee Retirement Income Security Act (ERISA) and the Internal Revenue Code (IRC) aims to eliminate unnecessary plan requirements for unenrolled participants in individual account plans and defined contribution plans. An unenrolled participant is defined as an employee who is eligible to participate in the plan, has been furnished the summary plan description and any other notices related to eligibility, is not participating in the plan, and satisfies criteria determined by the Secretary of Labor. The section states that, with respect to these types of plans, no disclosures, notices, or other plan documents (other than an annual reminder notice and any document requested by the participant) shall be required to be furnished under ERISA or the IRC to unenrolled participants.

Self-Certifying Hardships – Section 312

ChatGPT: This passage is amending sections of the U.S. tax code (specifically, sections 401(k)(14), 403(b)(7), 403(b)(11), and 457(d)) to allow plan administrators (such as those of 401(k) plans, 403(b) plans, and 457(b) plans) to rely on written certifications by employees or participants when determining whether a distribution from the plan is a hardship distribution. The certifications must state that the distribution is for an “immediate and heavy financial need” that is deemed appropriate by regulations set by the Secretary of the Treasury, that the distribution does not exceed the amount required to meet the need, and that the employee or participant has no other means to meet the need. The Secretary of the Treasury may also provide for exceptions to this rule if the plan administrator has actual knowledge that the certification is not accurate, and for procedures to address cases of employee or participant misrepresentation. The amendments made by this section will apply to plan years beginning after the date of the Act’s enactment.

Now here’s where it gets even more interesting, Section 603 on Catch-Up contributions. This is the summary ChatGPT produced:

ChatGPT: This is a section of the H.R. 2617 bill. It requires eligible participants whose wages for the previous year exceed $145,000 to make additional elective deferrals as Roth contributions if their employer-sponsored plan allows it. The Secretary of the Treasury may allow participants to change their election if their compensation exceeds the limitation after the election was made. The $145,000 amount will be adjusted annually starting in 2025 to account for cost-of-living increases. The changes made by this section will take effect for taxable years starting after December 31, 2023.

This is interesting because it raises valid concerns regarding AI, principally accuracy and reliability. Us retirement nerds know, as Brian Graff, CEO of the American Retirement Association (ARA), pointed out recently, there is a “significant technical error” with this provision of SECURE 2.0; in short, it essentially deletes subparagraph (C) of IRC §402(g)(1) which therefore eliminates the ability to make any pre-tax catch-up contributions.

The article explains that the Treasury Department has acknowledged that it appears to be a technical error. However, as you can see, ChatGPT does not mention this in its summary. Further, I decided to ask ChatGPT, “Are there any technical problems with this provision?” and this was the response:

ChatGPT: As a language model, I don’t have information on specific technical problems with this provision, but you can consult with legal or financial experts to determine if there are any potential issues or challenges with its implementation.

I think this sums up a number of considerations regarding AI. It highlights the compelling potential and reveals that humans still play a key role in the creative process. With respect to that process, I could observe my continued evolution with AI.

In the prior example of Section 603, I knew to ask, on that provision, if there were technical problems—and that changed my perspective to some extent—with respect to how I think about phrasing or definitions when looking to leverage AI for efficiency gains (as well as the value I place on the quality of the output, in certain situations). Lastly, just for fun, courtesy of ChatGPT here is this entire article, summarized for a 2^nd grader who is also pretending to be a pirate:

ChatGPT: Arrr matey! This be talkin’ ’bout makin’ it easier for sailors t’ save their loot for when they be retirin’. The captain of their ship should be takin’ a bit o’ their pay and stashin’ it away for them, unless the sailor says they don’t want them t’. And as time goes on, the captain should be savin’ a bit more. This way, the sailor will have more loot saved up when they hang up their hook. Ye savvy?

The SECURE Act 2.0 provisions are complex and we recommend discussing them with your Aldrich Wealth Advisor to determine if there are any strategies you should undertake now or in the future. As more details surrounding the SECURE Act 2.0 become available, we will continue to provide relevant updates and offer additional planning opportunities to help you achieve your financial and wealth transfer goals.