The Tax Cuts and Jobs Act (TCJA) doubled the federal estate, gift, and generation-skipping transfer (GST) tax exemptions at the end of 2017. Those provisions are scheduled to sunset at the end of 2025, returning the exemption to its pre-TCJA level (adjusted for inflation) — roughly $7 million per person, down from $13.99 million in 2025.
The sunset is the largest scheduled change to estate planning since the 2010 carry-over basis fiasco. Engines that handle estate planning have to model both the pre-sunset and post-sunset rules, the transition logic between them, and the strategies (gifts, GRATs, sales-to-IDGTs, SLATs, dynasty trusts) that depend on the exemption level. The dynasty-trust path in particular leans on the GST exemption — see generation-skipping transfer tax modeling for the inclusion-ratio mechanics that determine whether a sunset-era allocation locks in tax-free or tax-exposed status. This article is the working note on what's changing, what isn't, and what a 2026-grade estate-planning engine has to do.
What the sunset actually does
The sunset reverts several specific TCJA provisions to their pre-2018 form. The relevant changes for estate planning:
| 2025 (pre-sunset) | 2026 (post-sunset, projected) | |
|---|---|---|
| Federal estate tax exemption | $13,990,000 | ~$7,000,000 (estimated; depends on cumulative inflation) |
| Federal gift tax lifetime exemption | $13,990,000 (unified with estate) | ~$7,000,000 (still unified) |
| GST tax exemption | $13,990,000 | ~$7,000,000 |
| Top federal estate / gift / GST rate | 40% | 40% (no scheduled change) |
| Annual gift tax exclusion | $19,000 per donee (2025) | Inflation-indexed; no scheduled change |
| Step-up in basis at death | Available for all assets | Available; no scheduled change |
| Portability of unused exemption (DSUE) | Available; surviving spouse can use deceased's unused exemption | Available; no scheduled change |
The exemption halving is the headline. Most other estate-tax mechanics (rate, portability, step-up) don't change. But the halving has substantial planning implications because of how the IRS treats prior gifts at the time of death.
The "claw-back" question
The biggest open question through 2018–2024 was whether gifts made before sunset under the higher exemption would be "clawed back" into the estate at death after sunset. The IRS resolved this in 2019 with proposed regulations and finalized them in 2024: gifts made under the higher exemption are not clawed back. A donor who gifted $13M in 2024 against the $13.61M exemption and dies in 2030 with a $7M exemption keeps the benefit of the higher exemption for the prior gift.
The implication is that gifts made before sunset effectively "lock in" the higher exemption — but only for the gifted amount. The donor's remaining exemption at death is the post-sunset level minus prior gifts that exceeded the post-sunset level.
exemption_available = max(0, current_exemption − prior_taxable_gifts) + max(0, prior_taxable_gifts − current_exemption_at_gift)- current_exemption
- = Exemption at year of death
- prior_taxable_gifts
- = Total lifetime taxable gifts (above annual exclusion)
- current_exemption_at_gift
- = Exemption when each prior gift was made
What's wrong with most engines
We've audited the estate-planning logic in several wealth-tech and family-office tools. The recurring issues:
Hard-coded exemption. The exemption is in a constants file that hasn't been updated for the sunset. The engine produces estate-tax projections that assume $13.99M exemption in 2030 — wrong by ~$7M of taxable estate.
Single-snapshot exemption. The engine treats the exemption as a single value, not as a time series. It can't compute the no-claw-back protection for prior gifts.
Missing GST tracking. GST exemption is separate from estate exemption (though usually the same number). The engine treats them as the same field, breaking when generation-skipping transfers happen (multi-generation trusts, direct-skip gifts to grandchildren).
Portability gaps. The engine knows about the unused exemption from the deceased spouse but doesn't track the timely-filed Form 706 requirement. A surviving spouse who didn't file on the deceased's death loses portability — and the engine doesn't surface this as a risk.
Strategy-naive. The engine projects estate tax under "do nothing" but doesn't model the strategies that mitigate it — GRATs, sales to IDGTs, SLATs, ILITs, charitable remainder trusts. Strategy modeling is the actual planning value; tax projection alone is half the product.
The strategies that depend on the exemption level
Most estate-planning strategies have a "use it or lose it" dynamic at the sunset. The major ones:
- Strategy 1Outright lifetime giftsMake taxable gifts in 2025 to use the higher exemption. Locks in the higher exemption per the no-claw-back rule. Risk: donor needs to retain enough wealth for own needs.
- Strategy 2Gifts to ILITsFund irrevocable life insurance trusts in 2025 with significant Crummey gift contributions. The ILIT removes life insurance proceeds from the estate. Pre-sunset funding maximizes the lifetime contribution.
- Strategy 3SLATs (spousal lifetime access trusts)Fund a SLAT in 2025 from one spouse to a trust benefiting the other spouse. Locks in the higher exemption while preserving access for the non-donor spouse. Risk: divorce or beneficiary-spouse death.
- Strategy 4GRATsGrantor-retained annuity trusts. Less sunset-sensitive (GRAT economics depend on the §7520 rate, not the exemption). Continue to be useful pre and post sunset.
- Strategy 5Sale to an IDGT (intentionally defective grantor trust)Sell appreciating assets to an IDGT in exchange for a promissory note. Removes the appreciation from the estate. Pre-sunset, the seed-gift portion uses the higher exemption.
- Strategy 6Dynasty trusts (GST)Fund multi-generation trusts allocating GST exemption in 2025. The GST allocation lasts forever; pre-sunset funding allocates more GST exemption.
- Strategy 7QPRTs (qualified personal residence trusts)Transfer a residence to a trust with retained right to live there for a term. Less sunset-sensitive but still on the planning menu.
Each strategy has its own engine code path. An engine that models only outright gifts can serve a fraction of the planning population — usually the simplest cases. The HNW households that benefit most from sophisticated planning need engines that model the full strategy menu.
What a 2026-grade engine does
The properties we look for in a production-grade estate-planning engine:
2026 estate-planning engine invariants
- Track exemption as a time series, not a constant. Allow user override of future-year assumptions.
- Track GST exemption separately from estate / gift exemption, even when they currently have the same value.
- Model the no-claw-back rule for prior gifts. Compute exemption available at death from gift history.
- Track DSUE (deceased spousal unused exemption) with the Form 706 timely-filed requirement.
- Model state estate-tax separately. State exemptions are typically much lower than federal (e.g., MA at $2M, OR at $1M, WA at $2.193M).
- Model major lifetime gift strategies: outright gifts, ILIT funding, SLATs, GRATs, IDGT sales, dynasty trusts, QPRTs.
- Surface 'use it or lose it' decisions for clients with wealth above the post-sunset exemption.
- Project estate tax under multiple scenarios: do nothing, partial use of higher exemption, full use of higher exemption. Compare after-tax wealth at horizon.
- Account for step-up in basis when comparing gift-vs-bequest strategies. The step-up at death is a meaningful tax savings that lifetime gifts give up.
- Track beneficiary structures multi-generationally. The GST exemption is the engine's signal that a gift to grandchildren isn't double-taxed.
The invariants together represent a meaningful engineering investment. We estimate 4–6 engineer-months from a position-only retirement-projection engine to a full estate-planning engine, plus ongoing maintenance as state estate-tax laws change.
What this means for synthetic test data
Test data for estate-planning engines has to include:
- Households with net worth above and below both the $13.99M (current) and $7M (post-sunset) exemption thresholds
- Households in the "in-between" band ($7M–$14M) — where the sunset most directly affects them
- Households in states with state estate tax (MA, OR, WA, CT, RI, NY, MN, IL, DC, MD, ME, HI, VT)
- Beneficiary structures spanning multiple generations
- Lifetime gift histories, including pre-sunset gifts that will need no-claw-back protection
- Trust structures: ILITs, SLATs, IDGTs, GRATs at various stages of their lifecycle
- Married couples with portability planning (one spouse pre-deceased, DSUE elected or not)
- Charitable structures: CRATs, CRUTs, CGAs, DAFs
A general-purpose synthetic corpus without estate-planning depth cannot exercise this engine. The corpus has to be built with estate edge cases as a design goal, not as accidental coverage.
Key takeaways
- The TCJA estate exemption sunsets at end of 2025, halving from ~$14M to ~$7M per person. Most estate-planning engines have not modeled the cliff correctly.
- The IRS finalized the no-claw-back rule in 2024: pre-sunset gifts using the higher exemption don't reduce the post-sunset exemption below zero. Engines have to compute this per-gift with the exemption schedule tracked over time.
- Strategies like SLATs, ILIT funding, GRATs, and IDGT sales all interact with the exemption sunset. Engines that model only outright gifts miss most of the planning menu.
- State estate tax adds a parallel regime with much lower exemptions. Engines that model only federal estate tax produce wrong results for households in MA, OR, WA, etc.
- A 2026-grade estate-planning engine tracks exemption as a time series, models the no-claw-back rule, supports the major lifetime-gift strategies, and accounts for step-up basis trade-offs.
- Synthetic test data needs households across the exemption-band spectrum, in state-estate-tax jurisdictions, with multi-generational beneficiary structures and existing trust positions.