EU recall trends
Measured on 47,394 alerts, 2005 to week 42 of 2026

EU product recalls doubled since 2022, and one notifying country explains 45% of it

Alerts in the European Commission's Safety Gate rose from 2,099 in 2022 to 4,703 in 2025, after a decade stable around two thousand a year. The rise is not general: cosmetics account for 58% of it and Italy alone for 45%, and those are the same alerts. In 2026, at comparable weekly cadence, all three curves turn back down together.

This page states a measurement, not a verdict on product safety. An alert records an authority's action, not an accident. The four reserves that bound this finding are set out below, and the queries that produce every number are published so a third party can rerun them.

The finding

How many alerts, and in which years?

Every alert published in the Safety Gate weekly reports, counted by the year of the report that carried it.

EU Safety Gate alerts per year. 2026 covers 42 weekly reports of 52, so it is not comparable as an annual total.
YearAlertsYearAlerts
20192,15320233,341
20202,21320244,066
20212,10420254,703
20222,09920263,066 over 42 weeks

Which category carries the rise?

Cosmetics went from 52 alerts in 2021 to 1,708 in 2025, a factor of 33. On their own they are 1,502 of the 2,604 additional alerts between 2022 and 2025, that is 58% of the entire rise. In 1,677 of those 1,708 cases, or 98.2%, the risk recorded is chemical.

And which country notified them?

Italy went from 41 notifications in 2021 and 51 in 2022 to 1,227 in 2025, a factor of 24. That single shift is 1,176 of the 2,604, or 45% of the EU-wide rise. And 1,130 of Italy's 1,227 notifications in 2025 are cosmetics: the two findings are one finding.

The control that matters

Does the finding hold when the trend reverses?

This is what separates a measurement from a coincidence. If Italy and cosmetics carry the rise, their reflux must pull the total down with them. It does, and the three curves turn together.

Compared at weekly cadence rather than as annual volumes, because 2026 carries only 42 reports of 52 and an annual total would compare two different durations.

Alerts per weekly report, 2025 against 2026 to week 42.
20252026Change
All categories90.473.0-19%
Cosmetics32.812.8-61%
Notified by Italy23.67.0-71%
A side effect worth stating

Did China's share rise with the alerts?

No. It fell during the rise, which is the opposite of what "recalls are doubling" invites a reader to assume.

Share of alerts whose recorded country of origin is China.
Year2019202020212022202320242025
Share52.6%49.3%48.4%50.3%36.5%40.3%42.9%

The mechanism is measured, not assumed: the cosmetics that inflate the total are mostly European. Of the 1,708 cosmetics alerts in 2025, the recorded origin is Italy 598, unknown 264, Germany 144, France 129, Spain 102, and China 83. An influx of European alerts mechanically dilutes China's share of the denominator.

China remains by far the first country of origin in level: 24,279 of 47,394 alerts since 2005, or 51.2%, more than all other countries combined. The finding above is about the trend. Reading it as a statement about the level would be the exact error this page exists to prevent.

Bounds

What does this finding not say?

Four reserves. They matter as much as the numbers, and they are stated before anyone has to ask.

It is not a signal that products became less safe

An alert records an authority's action, not an accident. The rise measures control activity, and its concentration in one category and one notifier makes "product safety is deteriorating" unsupportable on this data.

No cause is asserted for the Italian campaign

It begins in 2023, therefore before the EU General Product Safety Regulation became applicable on 13 December 2024. That regulation cannot explain it. Saying why Italy changed pace would require an Italian source we do not have.

"Notifying country" is not "country of sale"

It is the authority that raised the alert. A product notified by Italy may circulate anywhere in the single market.

2026 is incomplete

42 weekly reports of 52. Everything about 2026 on this page is given at weekly cadence and never as an annual total.

Method

How can this be rerun?

The source is public and the queries are published, so a third party holding the same source can reproduce every number on this page.

Source. The European Commission's Safety Gate weekly reports since 2005, reproduced unchanged in our archive. Reuse under Commission Decision 2011/833/EU.

Scope. 1,121 reports and 47,394 alerts, from the first report of 2005 to report 42 of 2026.

Double-count control, run before publishing any figure. The table holds 47,394 rows for 47,394 distinct case numbers, and the 1,708 cosmetics alerts of 2025 are 1,708 distinct cases. No alert is counted twice.

Category control. Cosmetics is a single clean value, 6,236 alerts in total. The only other value containing the word is Other - Cosmetic bag, two alerts, unrelated.

The queries
-- alerts per year
SELECT r.report_year, COUNT(*) FROM alert_raw a
  JOIN report_raw r ON a.report_reference = r.reference GROUP BY 1 ORDER BY 1;

-- the category carrying the rise
SELECT a.category,
       SUM(CASE WHEN r.report_year = 2022 THEN 1 ELSE 0 END) AS y22,
       SUM(CASE WHEN r.report_year = 2025 THEN 1 ELSE 0 END) AS y25
  FROM alert_raw a JOIN report_raw r ON a.report_reference = r.reference
 WHERE r.report_year IN (2022, 2025) GROUP BY 1 ORDER BY (y25 - y22) DESC;

-- the notifying country carrying the rise
SELECT r.report_year, COUNT(*) FROM alert_raw a
  JOIN report_raw r ON a.report_reference = r.reference
 WHERE a.notifying_country = 'Italy' GROUP BY 1 ORDER BY 1;

-- the double-count control, to run before publishing any figure
SELECT COUNT(*), COUNT(DISTINCT case_number) FROM alert_raw;

The archive behind these numbers is the same one that answers our EU product recall API. The analysis was produced by automated measurement over that archive and reviewed before publication; the reserves above are part of the result, not a disclaimer appended to it.