Worldwide vehicle recall register: what was filed, and what was changed afterwards

Three governments publish vehicle safety recalls, and all three overwrite their own files. When NHTSA revises how many vehicles a campaign covers, when Transport Canada rewrites a remedy, when DVSA corrects a defect description, the earlier version is simply gone. No authority publishes what a recall said before it was changed, and no one else keeps it either.

Looking for one specific car? Check whether your own recall has been rewritten: type a make, model and year and the archive answers for that vehicle.

Want to check us rather than believe us? Download the dated wording manifest: a SHA-256 fingerprint of what every tracked US recall says right now. Keep the file and you can prove a later rewrite yourself.

This page keeps it. Every recall below is held with the date it was read, and when a figure moves the earlier figure stays here beside the new one.

Since 17 August 2026 we also keep the words. Every American notice is re-read on a six-hour cycle, and when the wording of a remedy, a summary or a consequence is rewritten, both versions are kept: the one we read before, and the one that replaced it. This paragraph used to say we were not yet storing the previous text. That is no longer true, and the proof is directly below.

Caught being rewritten

Removed words are struck through on the left; added words are marked on the right. A rewrite is not by itself wrongdoing: an authority correcting itself is an authority working. What is missing everywhere else is that the earlier wording simply disappears.

Recall wording is published by NHTSA. The comparison between one reading and the next is CarsMultiverse. A rewrite is not by itself wrongdoing: an authority correcting itself is an authority working. What is missing everywhere else is that the earlier wording simply disappears. These are downloadable: CSV or JSON.

The registers, side by side

What changed after it was published

This is the part that exists nowhere else. A recall is filed, the press reports the number, and weeks later the number is different, or the remedy is, or the list of affected models.

How to read the table. It is the full change log for all three countries. The What changed column tells you which kind of entry you are looking at. Rows marked remedy changed, summary changed or units changed are amendments the authority made after the recall was published: the thing this register exists to record. Rows marked first seen or new recall are the baseline: the moment a recall entered our holding, with nothing yet to compare it against. We show both rather than only the amendments, because the date we first read a record is what makes the later comparison checkable.

Every row that comes from a United States campaign carries its own anchor, so a link from our feed or a citation lands on that row rather than at the top of the table.

Cleared, 16 August 2026. Entries recorded before 14 August included rows where the authority had merely reordered the affected-model list rather than changed it. Those were not real amendments. They have been removed, and model lists are now compared as sets, so ordering alone can no longer register as a change. Checked again on 16 August: no such rows remain, and no rows remain that record only the authority’s own update date moving.

Check a car

If you have found a recall and are not sure how to read it, start here.

Search the United States register by manufacturer, model or campaign number. Canada and the United Kingdom are listed further down; for a specific vehicle in those countries, use the authority’s own VIN lookup, linked under each source below.

The United States in detail

NHTSA campaigns are read one at a time and rechecked for amendments, which is why the US section carries the interim do not drive and park outside warnings and the over-the-air remedy flag. Those fields are the ones most likely to be revised after filing.

Source: the US National Highway Traffic Safety Administration recalls API. Recall text is the manufacturer or agency wording as published by NHTSA. US Government public domain.

Canada: the most recent recalls

Transport Canada rebuilds a rolling sixty-day file every day, listing every recall it has touched in that window. A recall from 2024 reappearing in today’s file means the agency has just changed something in it. The final column is the authority’s own last-update date, so you can see the gap between when a recall was published and when it was last altered.

That is the recent slice. The whole Canadian register, every notice Transport Canada has published, is at Canadian recalls.

United Kingdom: the most recent recalls

DVSA publishes a single file covering every recall since 1992 and replaces it in place. It carries no update date of any kind, so the only way to know a British recall has been revised is to have kept the previous copy.

That is the recent slice. The whole British register, every notice back to 1992, is at UK recalls.

Why the changes matter

A recall is a safety instruction, and safety instructions get corrected. Sometimes the correction is administrative. Sometimes the number of affected vehicles rises by an order of magnitude, or an interim instruction to stop driving the car is added, or quietly removed. An owner who read the first version and never came back has the wrong information and no way to know it.

Holding the earlier version is not a criticism of any of these agencies. They revise campaigns because manufacturers give them better information, which is how the system is meant to work. The gap is that the revision is invisible, and that is what this register closes.

What each authority does when a recall grows

A recall rarely stays the size it was announced. What differs between countries (and this is not written down anywhere else we can find) is how the growth is handled. It is the reason this register covers three countries rather than thirty: amending a published recall in place turns out to be a rare habit, not a normal one.

Two different ways of hiding the same thing. Where a recall is edited in place, the correction is invisible because the earlier version is gone. Where a second recall is filed instead, the correction is invisible because nothing joins the two records together. This register exists to close the first gap, and the second is what it will do next. Read the full study, with sources and what we could not establish.

Follow the record

The register reads the authorities on a schedule and updates itself. Three ways to be told rather than to keep checking:

• A million recalled cars with no repair on record: 18% of Dutch recall obligations have no repair logged against them.
• The same recall, two countries, two stories: matched across borders by the maker’s own code, months apart, and described with different severity.
• How long before anyone tells you your car is unsafe? The median gap between a maker declaring a defect and owners being told is 18 days; the worst on record is ten years.
• RSS feed: every new and amended United States recall as it is observed. Point a reader at it and you see a revision the day we do.
• What changed this week: the written weekly summary, generated from the same data.
• The changes as JSON: every revision observed across all three countries, for anything you want to build.

Using this professionally

If you are writing about a recall, checking a claim, or building on this data, start here: for journalists and researchers: what we hold, the machine-readable feeds, how to cite a live dataset, and what we deliberately will not claim.

How to cite this register

This is a live dataset: it changes when the authorities change their records. A citation therefore needs the date you read it, not just a URL: the same entry may read differently in a month, which is the whole point of the register.

CarsMultiverse (2026). Vehicle recall change register: what was filed, and what was changed afterwards. Read [date]. https://carsmultiverse.com/vehicle-recall-change-register/

If you are citing a specific revision, quote the campaign number and the date we observed the change: both appear in every row and in the JSON. That lets anyone check the authority’s current record against what it said when we read it.

The recall wording belongs to the issuing authority. The dated comparison between versions is ours, and it is free to reuse with a link back. We would rather be checked than believed.

What this register holds, exactly

Precision matters more here than a good line, so:

The gap is the previous text of a rewritten remedy or summary. Capturing it is a change to how the register stores its readings, and it is the next thing to be built. Until it is, this page will not claim to have it.

What is already true, and is not published anywhere else: no authority in this register tells you that a recall was revised at all. Recording the fact and the date of every revision is itself the thing nobody keeps.

What a recall used to say

Until today this register could tell you that a remedy was rewritten. It could not tell you what it said before, because NHTSA replaces the text in place and nobody was keeping the earlier version. From 17 August 2026, we are.

Every few minutes we re-read the wording of the most recently published US campaigns and compare it, word by word, against the wording we last held. When it differs we keep both versions, each with the date we read it. The authority publishes only the later one.

Two honest limits. First, this history begins on 17 August 2026: the earlier wording of revisions recorded before that date is gone, and no amount of work recovers it. Second, we now track the entire US holding: every campaign in the register, not a recent slice of it. It began at 100 on install day, because the public JSON feed caps its recent list at that number; reading the register’s own index instead lifted it to the full set the same afternoon.

The comparison is ours. The wording is NHTSA’s, and we say so on every diff.

France and Poland, from today

Two more authorities are now being read on the same terms as the others. France: the DGCCRF’s RappelConso feed, vehicles category. Poland: the Transportowy Dozór Techniczny register of service actions. Neither publishes a revision history, so the comparison has to be ours: every notice is hashed on the day it is read, and when the hash moves while the reference does not, that is an amendment nobody announced.

We also hold 553 recall records from France and Poland, each hashed on the day it was read, with 500 amendments recorded since we began.

France: the most recent recalls we hold

20 records held from DGCCRF, RappelConso. Showing the 20 most recent. The first read by us date is what makes a later change checkable.

Poland: the most recent service actions we hold

533 records held from Transportowy Dozór Techniczny. Showing the 20 most recent. The first read by us date is what makes a later change checkable.

French and Polish notices changed after publication

Both were baselined on 17 August 2026. There is nothing to show yet, and saying so is the point: a register that reported changes on day one would be reporting its own arrival, not theirs.

Where a revision can be seen at all

Three countries are in this register because three countries publish enough for a revision to be visible. Most do not. We keep a public map of every major car market: who publishes, what they publish, and whether an amendment can be seen by anyone outside the agency: vehicle recall transparency by market. Eight of the largest markets are effectively closed, and in several of them a recall can be amended with no way for the public to know.

What this page will not do

It will not tell you whether your particular car is affected: only the manufacturer and the authority can confirm that from your VIN. It is not a substitute for acting on a recall notice, and it carries no advice about whether a vehicle is safe to drive. Recall wording here is the authority’s own text, unedited; where a figure is ours, it is described as such.

How this page is made

Three free public sources, read on a schedule and compared with the copy held here:

• United States: the NHTSA recalls API. US Government public domain.
• Canada: Transport Canada’s Vehicle Recalls Database open-data file, rebuilt daily. Open Government Licence: Canada. VIN lookup at recalls-rappels.canada.ca.
• United Kingdom: the DVSA recalls file, every campaign since 1992. Crown copyright, Open Government Licence. Vehicle lookup at check-vehicle-recalls.service.gov.uk.

Nothing on this page is typed by hand. When a source changes, the page changes; when a source stops responding, the last good copy stays and the run is abandoned rather than writing a damaged record. Each country’s figures show the date they were last read.

Two authorities that cannot join this register. The European Safety Gate is held here, on its own page, but its alerts carry no amendment field, so a rewrite in Brussels cannot be seen and its rows are not part of this change log. Germany’s KBA publishes no machine-readable file at all. Both are noted here rather than quietly omitted.

Take the data

All of it is open, and you are welcome to it:

• vehicle-recalls-worldwide.json: what each authority has published and what we hold
• vehicle-recall-changes-worldwide.json: every revision observed, all three countries

Attribution: recall wording belongs to the issuing authority under the licences above. The dated comparison between versions is ours; please link back if you use it.

Every figure on this page comes from a table you can have. 99 rows, CSV JSON All 14 datasets