What a banning order is
The NDIS Commission’s enforcement register records six kinds of action. Compliance notices are far and away the most common, at 2,358 of the 3,668 rows on the register today. Banning orders are the second largest category at 789, and they are different in kind: a compliance notice tells a provider to fix something, while a banning order stops a person or an entity from providing supports at all, in whole or in part.
Because the Commission publishes these as a statutory function, the register is a public record rather than an allegation. This story reports only its aggregate shape. We have not named anyone on it, and nothing here should be read as a claim about any individual.
The duration finding
| Term | Orders | Share |
|---|---|---|
| No end date recorded | 316 | 40% |
| Under 2 years | 145 | 18% |
| 2 to under 5 years | 268 | 34% |
| 5 to under 10 years | 37 | 5% |
| 10 years or more | 23 | 3% |
Source: NDIS Commission compliance and enforcement register, full CSV export, downloaded and parsed 28 July 2026. Shares are rounded and sum to 100 per cent. Term is calculated from the register’s own "date effective from" and "date no longer in force" columns.
The headline number is the 316. Two in five banning orders on the register carry no recorded end date, which on the face of the data means they run until something changes them. The shortest time-limited order runs about five weeks; the longest runs twenty years. In between, the distribution is tight and unmistakably clustered: 268 of the 473 time-limited orders, a clear majority, run between two and five years, and the median sits at exactly three.
That clustering is the interesting part. It suggests a working convention rather than a case-by-case calculation, with a three-year order as the ordinary response and the indefinite order reserved for a different class of matter. The register does not say why any particular term was chosen, and we are not going to invent a reason.
The finding that matters more: who these orders are against
Of the 789 banning orders, 570 have no ABN recorded. Only three of those 570 names contain any corporate marker at all, and 90 per cent are two or three words long. The remaining 219 orders do carry an ABN, and just under half of those names contain a corporate marker, with the rest looking like sole traders.
The reasonable reading, which we state as a reading rather than a fact on the register: the banning power is overwhelmingly used against individual people working in the scheme, not against companies. That is worth sitting with, because most public argument about NDIS integrity is conducted in the language of providers, markets and business models. On the Commission’s sharpest instrument, roughly seven in ten subjects appear to be a person rather than an organisation.
The growth, with the base rate attached
| Year | Orders |
|---|---|
| 2019 | 7 |
| 2020 | 15 |
| 2021 | 23 |
| 2022 | 54 |
| 2023 | 105 |
| 2024 | 134 |
| 2025 | 280 |
| 2026, to 28 July | 171 |
The temptation here is a surge headline, and it would be wrong. 2025 more than doubled 2024, which is a genuine step change. But 2026’s 171 orders cover slightly under seven months; at that rate the year lands near 295, which is a modest increase on 2025 rather than another doubling. The escalation happened in 2025 and 2026 is so far holding roughly at the new level. That is a less dramatic story than the raw column suggests, and it is the accurate one.
By state, the orders fall NSW 293, Victoria 184, Queensland 127, South Australia 94, Western Australia 62, Northern Territory 11, Tasmania 10 and the ACT 4, with four blank. Those are raw counts and we have deliberately not built a league table from them: they are not adjusted for population, for the number of participants, or for the number of registered providers in each state, and any of those adjustments could reorder them.
A caution about the register itself
One thing we learned by downloading the register twice. On 27 July it carried 3,671 rows. Today it carries 3,668. Every category is identical except compliance notices, which fell by exactly three.
The register is not append-only. Rows leave it. That matters for anyone citing a total from it, including us: a count is a snapshot of a live system, not a cumulative historical tally, and two people quoting it a day apart can both be right and still disagree. It is also why every figure in this story carries the date we pulled it. We do not know why those three rows went, and the register does not say.
What we would want to know next
Three things the register cannot answer on its own. Whether an order with no end date is formally permanent or simply open-ended pending review. Whether the three-year clustering reflects a published policy or an internal convention. And what happens after a time-limited order expires, given that the earliest ones are now reaching their end dates: does the person return to the workforce, and does the Commission track that. Those are questions for the Commission, and they are the ones we would put.