A breach notification count is not a count of affected people
Distinguishing notifications, incidents and people helps readers understand what a breach report can and cannot tell them.
Explain harm without changing the counting unit
Breach statistics can inform incident response and data-minimisation discussions. A number of notifications and a number of affected people nevertheless describe different things. Mixing them can mislead the people trying to understand the report’s scope. Truthound can check a reporting table that retains these distinctions. It does not identify actual victims, determine individual harm or prevent a disclosure of personal information.
What the number 667 counts
On 26 October 2026, Australia’s OAIC released Notifiable Data Breaches statistics for January 2026 through June 2026. It described 667 notifications, increases in large-breach bands and an updated dashboard. That is not a census of all breaches or a count of mutually distinct affected people. The publication date and notification reporting period also need separate treatment when the figures are reused in another reporting system.
Record what each quantity measures
A proposed table keeps notification identity, reporting period, affected-person band, cause class and sector separate. A notification’s size band differs from the count of notifications within a sector. The same person may appear in more than one incident, so adding bands does not produce a unique-person total. Keep a value unknown where the available publication does not establish it, rather than manufacturing a complete-looking measure.
Minimisation requires additional evidence
A large breach can justify reviewing what an organisation collects and retains. The aggregate statistics do not determine which personal fields a particular service should delete. Its owner must separately examine collection purpose and retention basis. Public explanations should also distinguish a notification trend from the route for confirming an individual’s exposure, so that a statistical summary does not silently replace information needed to exercise personal rights.
Check only the declared reporting contract
Truthound Core Python SDK 3.3.4 (open-source SDK, Polars backend) documents ColumnPairInSetValidator for operator-approved category pairs. It could check reporting-item and size-band combinations. MultiColumnUniqueValidator can test a specified notification key inside a defined reporting scope. Declare periods and null handling first. A unique key does not establish that the underlying incident is unique or that the affected people never appear in another notification.
An existing aggregation contract can remain
Pandera 0.35.0 (MIT open-source Python DataFrame validation library; pandas, Polars and other supported backends) declarative and custom checks provide another route to the same classification contract. Retain a reporting workflow that already manages units and exceptions correctly. Compare it with Truthound at the table-admission stage, rather than as a breach detector or a way to identify individual harm. Neither tool’s capability documents establish that it discovers unreported incidents or people absent from the source statistics.
Do not substitute statistics for personal guidance
Start by placing the counting unit and period beside each reported measure. If there is no basis for a unique-person figure, say so and retain the relevant incident-level confirmation route. Revisit the explanation even after the categories pass if readers cannot understand what is being counted. Useful reporting distinguishes what people can know from the available evidence and what still requires another source of confirmation.
Conditions
Define notification count, affected-person band and reporting period before combining aggregates.
Alternatives
A Pandera 0.35.0 contract can retain equivalent category and counting-unit controls.
Limitations
Notifications are not unique people or a complete count of every real breach.
Evidence
- Event checked
- Subject checked
- Modified
