A 2026 national agreement funds disability supports outside the NDIS for one cohort — children aged 8 and under — and leaves the size, funding and start date for every other cohort undefined

What is happening?

On 2023-12-07 the independent review of the National Disability Insurance Scheme delivered its final report. The first of its 26 recommendations was to invest in foundational supports — a new layer of assistance for people with disability under 65 and their families, sitting outside the individualised budgets that the scheme funds. At the same time the National Cabinet agreed to bring annual growth in scheme costs down to 8 percent.

More than two years later that layer exists as a signed instrument. The Foundational Supports national agreement 2026-2031 was signed by all eight states and territories and by the Commonwealth. The Australian Capital Territory, the Northern Territory and Tasmania signed on 2026-02-02 and the remaining jurisdictions in January 2026; the Commonwealth signed on 2026-02-27. The agreement runs to 2031-06-30.

What the agreement actually funds and dates is one cohort: children aged 8 and under with developmental delay or autism or both and with low to moderate support needs, together with their families, carers and kin. The program is called Thriving Kids. Clause 6 describes it as the first stage of a staged rollout.

For everyone else the agreement records an absence rather than a plan. Clauses 6 and 49 to 51 provide that further cohorts are to be settled in separate bilateral agreements without amending the national agreement, and that the Commonwealth will supply four things at the earliest opportunity — a definition of the cohort, an estimate of its size, a budget quantum and a preferred start date. At signature none of the four had been settled for any cohort beyond the first.

The money is capped. Up to 10 billion Australian dollars over five years, split 50:50 between the Commonwealth and the states on a population basis. Of the Commonwealth share, at least 1.4 billion Australian dollars is allocated directly to states and territories for Thriving Kids, distributed on a population basis across ages 0 to 64.

Whose problem is this?

The party that must define the remaining cohorts is the same party that agreed to build the tier for them. Nothing in the agreement sets a date by which that definition has to arrive.

Where does this problem end?

The boundary here runs between a support tier that has been agreed and the cohorts for which that tier has not yet been defined. Spending restraint inside the scheme is the background to that gap and not the gap itself, and this document uses it only where a source names it as a cause.

What is the state now, and what should it be?

Now

What it should be

The target state is defined in kind and not in quantity. The review asked for a tier open to every person with disability under 65 who does not hold an individualised budget, and the agreement itself names the four things the Commonwealth must still supply for those people. Three of the four are quantities and the fourth is a date, so the agreement is a description of what the target state must contain rather than a statement of what it is.

How big is it?

The quantities that bound this problem are published. The quantity that measures it is not.

None of these quantities converts into a count of the people this document is about. They rest on different definitions, different base years and different age ranges, and combining them would produce a number that no source supports. The census figure is a self-report with no age limit and no relation to scheme eligibility. The participant count is an administrative total five years later. The planned reduction is a target rather than a measurement.

The term that would settle the count is the one term the agreement records as missing. Clause 51 has the Commonwealth supplying an estimated size for each further cohort at the earliest opportunity, and at 2026-08-15 no such estimate had been published for any cohort. The affected population is therefore recorded as not derivable, and the reason is not that research fell short but that the party responsible for the estimate has not produced it.

Under what conditions does it arise?

1. The tier was agreed in principle long before it was defined in detail. The recommendation arrived in December 2023 and the binding agreement took more than two years to sign. What was signed defines one cohort out of all the cohorts the recommendation named, so the interval between agreeing and defining did not close when the agreement did.

2. Addition without amendment removes the deadline. Clauses 49 to 51 let further cohorts be added by bilateral agreement without touching the national agreement, which means the instrument is complete as signed even though most of its intended coverage is empty. A schedule that can always be added later is a schedule that nobody is late on.

3. The funding is capped and the first cohort is first in line. With up to 10 billion Australian dollars over five years and at least 1.4 billion of the Commonwealth share committed to children aged 8 and under, every later cohort negotiates against the remainder rather than against the need.

4. Pressure inside the scheme runs on a clock and the tier outside it does not. Expenditure grew 11.3 percent against a target of 8 percent, a reform bill to tighten eligibility has a parliamentary timetable, and the number of people who will need a place outside the scheme rises while the place outside the scheme has one cohort in it.

5. Existing state services are held at a baseline. The additionality clause measures state and territory disability services against their 2022-23 level, so those services cannot be counted as the new tier and are not expected to grow into the gap either.

What has been tried?

Two directions have been pursued at once. One narrows the scheme on a legislative timetable, and the other builds the tier that people leaving the scheme would fall into, on no timetable at all for anyone over the age of eight.

What was found?

Why is it still unsolved?

Institutional gap — the tier that would close it has been agreed, signed and funded, and it covers one cohort out of all the cohorts it was agreed for.

The instrument is not missing. That is what makes this different from a promise that was never kept. A national agreement exists, all nine governments signed it, money is attached to it and a program under it will start taking children in October 2026. Everything that a reader would look for as evidence of action is present. What is absent sits inside the instrument, in the clauses that say the rest will be negotiated later.

The second part is that later has no edge. Clauses 49 to 51 allow further cohorts to be added by bilateral agreement without amending the national agreement, and the obligation on the Commonwealth is to supply the definition, the size, the budget and the start date at the earliest opportunity. An obligation phrased that way cannot be breached on any particular day. The agreement is therefore being complied with at every moment in which nothing happens.

The third part is that the money was capped before the coverage was defined. Up to 10 billion Australian dollars over five years is a ceiling set against a tier whose population has never been estimated in public. The first cohort has its allocation. Every cohort after it negotiates against what is left of a fixed number rather than against what the need turns out to be, which means that defining the cohorts later is also a way of deciding how much of the cap they can claim.

The fourth part is that the gap produces no record of itself. Nobody is refused foundational supports, because there is no place to apply and no eligibility test to fail. There is no queue, no waiting list and no rejection letter. A gap that generates no administrative artefact generates no number either, and a problem with no number is hard to put on any agenda against a scheme whose growth rate is measured every quarter.

What observation would mean it is solved?

Candidates — (a) bilateral schedules are published that name each remaining cohort with its size, budget and start date (b) services under the tier are counted as actually reaching people with disability under 65 who are outside individualised budgets (c) the number of people who fall between the scheme and the tier falls and keeps falling.

(a) alone counts paper. A schedule can define a cohort narrowly enough that the definition does more excluding than including, and a start date can sit beyond the term of the agreement that promised it. The first cohort shows how far a signed schedule can be from a delivered service, with a partial start eight months after signature and full national rollout nearly two years after that.

(b) is the observation that matters and the one nobody publishes. There is no count today of people receiving foundational supports outside the scheme, because for cohorts other than the first there is nothing to count. Until that series exists, coverage can only be inferred from spending, and spending against a capped allocation says how much was disbursed rather than how many people were reached.

(c) can move for the wrong reason. The number of people between the two systems falls if the tier arrives, and it also falls if scheme eligibility widens again, and it rises if the reform bill tightens eligibility while the tier stands still. Read alone it cannot distinguish a gap being closed from a gap being redefined, so it has to be read against (a) and (b) and against what the scheme did in the same period.

What is it connected to?

Fills with researchthe aged care system, which is where people over 65 with the same support needs are directed; state and territory disability services held to a 2022-23 baseline by the additionality clause; the early childhood education and health services that would deliver the first cohort program; and the scheme reform bill, which changes how many people arrive at this boundary. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• How many children the first cohort covers. No source opened here carries a government estimate of the number of children aged 8 and under who would be eligible for Thriving Kids, which is the one cohort that has money and dates attached to it.
• The size, budget and start date of every other cohort. This is not a silence in the sources but a silence the agreement records about itself, in clauses 49 to 51, and it is the reason the affected population in this document is not derivable.
• Any date attached to the four items the Commonwealth owes. The agreement says they will be provided at the earliest opportunity, and nothing opened here names a date by which they must arrive.
• How many current participants would face reassessment or lose eligibility under the reform bill. The departmental fact sheet and the parliamentary committee page both refused automated retrieval, so no official figure was confirmed.
• Whether the first cohort funding is 1.4 billion or about 4 billion. The agreement text opened here supports at least 1.4 billion from the Commonwealth share. Other reporting cites a larger figure that appears to combine Commonwealth and state contributions, and the difference could not be reconciled against a document.
• The modelling behind the advocacy figures. The claims of a 10 percent real funding cut, 94 million paid care hours and 51,600 jobs come from a disability representative organisation citing its own modelling, and that modelling document was not obtained.
• The originating document for the 8 percent target. The ministerial release that would carry the date and wording of the National Cabinet decision refused automated retrieval, so the target is reported here at second hand.
• The outcome of the parliamentary vote. The committee reported on 2026-08-14 and the vote falls after 2026-08-15, so no source could state a result.

See the evidence

One primary document was opened directly. The signed national agreement is the source for the signing dates, the term, the funding cap and the split, the allocation to the first cohort, and the clauses that defer every other cohort — which is to say that the central claim of this document rests on the text of the instrument itself rather than on any account of it. Five sources could not be opened at all, and four of the five are government pages, which is why the start dates for the first cohort program, the reassessment scale under the reform bill and the originating document for the 8 percent target are all reported here at second hand and graded accordingly. Where opened sources overlap they agree on the participant total and on the growth rate, both of which reach this document through the same quarterly summary rather than through two independent readings. Where they diverge the divergence is left visible: the allocation to the first cohort appears as at least 1.4 billion in the agreement and as a larger combined figure in other reporting, and the cost growth target appears as 8 percent in the 2023 decision and as 5 or 6 percent in later reporting that names no document. The claims about the effects of the reform bill come from a disability representative organisation and are recorded as its position rather than as findings. This is a Path A output, research-based definition, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range Not derivable

The reason and what is missing are listed under “What is missing” below