Raising the dependent care FSA limit to $7,500 for 2026 was the easy headline. Keeping the plan qualified under section 129 is where employers and TPAs still lose sleep — especially at companies where executives max the account while frontline staff elect little or nothing.
Treasury and the IRS have been working on that gap. On August 11, 2026, they published proposed regulations (REG-101355-26, also in IRB 2026-37) that finally walk through the four statutory nondiscrimination tests for dependent care assistance programs. Most of the press coverage focuses on employer Trump Account contributions; the dependent care FSA sections are the part cafeteria-plan administrators should read.
Written comments are due September 25, 2026 (Regulations.gov docket IRS-2026-0925-0001). A public hearing is scheduled for October 15, 2026, if speakers sign up by the same comment deadline.
The four tests — and what the proposal clarifies
Section 129 has always required more than “offer the benefit to everyone.” A dependent care assistance program must satisfy:
- Contributions and benefits — no discrimination in favor of highly compensated employees (HCEs) or their dependents.
- Eligibility classification — the group of employees who may participate must be reasonable and not discriminatory in favor of HCEs.
- Owner concentration — no more than 25% of total dependent care assistance for the year may go to more-than-5% owners (and related individuals).
- Average benefits — average benefits for non-HCEs must be at least 55% of average benefits for HCEs.
The proposed rules under § 1.129-2 translate those statutory phrases into operational steps: same-terms contribution design, classification tests modeled on qualified-plan § 1.410(b)-4 safe harbors, and a defined formula for the 55% average benefits calculation.
The 55% test: who counts, and when
This is the test TPAs see fail in practice when a handful of HCEs elect $7,500 while most NHCEs elect $0 because they do not use daycare or did not understand the election.
The proposal would measure average benefits as of the last day of the plan year, counting employees who received more than zero dependent care assistance during the year (salary reduction or employer-paid). Employees with $0 elections generally drop out of the denominator — which can help or hurt depending on your workforce mix.
Certain employees may be excluded from the eligibility and average benefits tests, including those under age 21 or without one year of service (subject to rules similar to section 410(b)(4)), and collectively bargained groups where dependent care was a good-faith bargaining topic.
If you run nondiscrimination testing today using informal spreadsheets, compare your method to the proposed § 1.129-2(d) framework before year-end testing season.
A correction path tied to Form W-2 timing
When a plan fails the average benefits test — or the 25% owner concentration test — employers often ask whether anything can be fixed after the fact without unwinding reimbursements.
The proposal would allow remediation through gross income inclusion for affected HCEs (and, for owner concentration failures, related adjustments described in § 1.129-2(j)), on or before the Form W-2 deadline for the year the benefits were provided. The regulations describe how to compute “excess benefits” and allocate reductions among HCEs when not everyone is over the threshold.
That is not a free pass to ignore testing until January. It is a proposed administrative relief valve — and it assumes payroll can still process taxable adjustments and issue correct W-2s. TPAs should confirm whether their platforms can support gross-up or taxable reversal workflows if a client fails late in the year.
What this is not
These are proposed rules, not final regulations. Until Treasury and IRS finalize and set applicability dates, existing section 129 law still governs — and many employers continue to rely on informal testing practices and plan document language written before the $7,500 limit.
The same NPRM also covers Trump Account employer contribution programs under section 128. That is a separate written plan with parallel nondiscrimination concepts. A dependent care FSA offered through a cafeteria plan is still tested under section 129, even if the employer also launches a Trump Account program elsewhere.
What employers should do this week
- Pull your 2025 plan-year testing file (or ask your TPA for it) and rerun the 55% and 25% owner tests using the proposed counting rules — especially if HCE election rates jumped after the limit increase.
- Review eligibility classifications in the cafeteria plan: “all full-time employees” is fine until it is not; narrow classes need the facts-and-circumstances or numerical safe harbor analysis described in the proposal.
- Decide whether to comment by September 25 if you see operational gaps — for example, mid-year hires, zero elections, or owner-employees with high reimbursements. Short, concrete examples from real payroll data help IRS more than generic support letters.
- Calendar the hearing (October 15, 2026) if you want trade groups or counsel to speak on remediation timing or TPA reporting obligations.
Dependent care FSAs only stay tax-free if the plan, not just the election form, passes muster. These proposed rules are the clearest map IRS has published in years — worth reading before open enrollment pushes another wave of $7,500 elections.