EU auditors say AI is helping tobacco smugglers pick their next target

The European Court of Auditors puts the annual loss at €13bn and admits the bloc has no map of the illegal market.


Petri Sarvamaa speaking, with European flags blurred in the background

Petri Sarvamaa, a member of the European Court of Auditors.

Image Credits Credit: © European Union 2022

The European Court of Auditors has published two findings that sit awkwardly together. The EU has no reliable picture of the size or structure of its illicit tobacco market.

The criminal networks running that market, according to the auditor who led the report, are using artificial intelligence to work out where to move next.

Special report 23/2026, published on Wednesday under the title “Combating the illicit trade in tobacco in the EU: fragmented efforts and persistent gaps”, estimates the annual cost to EU and national budgets at €13bn. Roughly one in 10 cigarettes made in the bloc is smuggled or produced illegally.

Petri Sarvamaa, the report’s reporting member, made the technology point in an interview with EUobserver.

“Artificial intelligence has hugely benefited the criminals because they can very easily find out where to move next,” he said, describing organisations that track enforcement activity across member states and analyse where the loopholes are.

Strip the acronym out and what he is describing is an information advantage. Smuggling has always depended on knowing which border is loosely staffed, which member state has the softest excise regime and which enforcement agency is looking elsewhere this month.

What has changed is how cheap it now is to assemble that picture from public sources and keep it current. The auditors’ other finding is what makes this uncomfortable:

“We do not have an EU-wide picture of the true size, structure, and economic impact of this illegal market,” Sarvamaa said.

One side of this contest has invested in data infrastructure. It is not the regulator.

What the report does not do is say how any of this was established, which is the gap a technology desk should notice.

Sarvamaa’s claim about AI came in an interview rather than in the audit, and from what is visible of the document it names no tools, no cases and no seized systems.

That does not make it wrong. Enforcement officials across Europe have been describing the same shift for a year.

But an assertion that criminal groups have operationalised AI is the kind of claim that hardens into received wisdom without ever acquiring evidence, and auditors are supposed to be the people who insist on the difference.

The scale is not theoretical. Belgian investigators found a factory running machines capable of about 1 million cigarettes an hour. The largest plant dismantled in Spain yielded 3 million counterfeit packets.

Alongside cigarettes, the report counts nearly 21,000 tonnes of illicit non-cigarette tobacco. The EU’s anti-fraud office prevented €178mn in lost tax revenue last year, which, set against a €13bn annual estimate, gives the ratio the auditors are complaining about.

Their diagnosis is institutional rather than technological. Coordination is fragmented, enforcement is uneven between member states, information exchange is inconsistent, and there is no harmonised EU law covering the trade.

Those are the conditions that make cross-border arbitrage worth analysing in the first place. An organisation that can see all 27 jurisdictions at once has an obvious advantage over 27 authorities that cannot see each other.

The product mix is shifting too. Vapes and heated tobacco now account for about 13% of the illicit market by value, a category that barely existed when the enforcement architecture was designed and one whose supply chains run through the same e-commerce and logistics rails as legitimate consumer electronics.

“That is €13bn that never reaches our schools, hospitals, or other public services,” Sarvamaa said of the revenue loss.

The number is worth treating with the same caution the auditors apply to it, since it derives from external studies rather than direct measurement, which is precisely the measurement problem the report is complaining about.

An institution that cannot size the market cannot size its losses either, and the honest version of this report is a request to be allowed to find out.

Get the TNW newsletter

Get the most important tech news in your inbox each week.

Published
Back to top