Society & safeguards

Hacked fridges across Europe: the hackers didn't need AI

They walked in through a door nobody had locked.

6 min read
Hacked fridges across Europe: the hackers didn't need AI

A Monday morning, 150 °C in the cold room

On Monday 2 November, Alex Pearce is doing his morning round at his wholesale fruit and veg business in London. On refrigeration unit no. 2, the defrost light is on and the display reads 150 °C. Same on unit no. 4. As he walks toward the biggest one, its light comes on too, he told the Guardian.

He thought of a cyberattack straight away, walked over to the box that connects his fridges to the internet and yanked out the cable. "If it had happened on a Saturday morning, when I'm not here, it could have ended in a fire," he said.

He wasn't alone that morning. Shopkeepers in Belgium, Estonia, Ireland, Finland, Luxembourg, the Netherlands and Sweden reported the same problem, with fridges and freezers pushed as high as 50 °C, according to the Guardian.

A month of overheating fridges

France got hit first. In the night of 11 to 12 October, around 700 of the country's 10,000 supermarkets had their refrigeration systems thrown off remotely, according to Finance Minister Roland Lescure, quoted by AP. The Mousquetaires group counted 300 of them among its Intermarché and Netto stores. The Paris prosecutor's office opened a case.

Then, on 2 November, it was the turn of cold rooms belonging to the Établissement français du sang, the French blood service, at three regional sites, through the remote monitoring software of its contractor Coolworld (per an AFP dispatch). According to the satirical weekly Le Canard enchaîné, rooms set to -30 °C were programmed to reach 110 °C. The service says "no loss of blood products is to be reported," thanks to its alarms and a technician on call.

The story taking hold: AI is opening the doors

One explanation caught on fast. Richard Horne, head of the NCSC, the UK's cyber security agency, says we are in the "foothills" of a new landscape. He was talking about the flaws AI is uncovering, so many that they could be said to be "waiting to be exploited."

The Guardian adds a figure: since the April release of Mythos, Anthropic's model that can spot flaws in bulk, the number found by big tech companies has reportedly been multiplied by eight. The paper doesn't say where the number comes from, and we couldn't find it anywhere. The closest public measure, from the research institute Epoch AI, counts serious flaws published by 21 large organizations: in June, more than 3.5 times the monthly record from before Mythos. The wave of flaws is real enough: Bank of America had to call in 1,000 consultants to fix what Mythos kept turning up.

Entrepreneur Azeem Azhar went as far as calling it a "Wuhan moment." The image says it all: AI as the source of an outbreak of hacks.

What the investigators say

But on the fridges, the people investigating tell a different story. "The key thing is that this was not an AI attack," said Paul Chichester, the NCSC's director of operations, quoted by AP. The attackers used vulnerabilities that had been "around for a very long time." He told the Guardian they "really didn't need" AI to find these already-known flaws, down to the default passwords printed in a manufacturer's manuals.

Nicole Carignan of the security firm Darktrace reads it the same way: no evidence that AI was used, and the techniques involved didn't require it. In her account, the attackers got in through boxes exposed to the internet, end-of-life software and factory passwords. The NCSC chief himself describes French fridges "connected directly to the internet with default passwords."

It's a bit like blaming a state-of-the-art lock pick for opening a house whose back door was never locked.

Control boxes nobody was updating

The hardware is known. As early as 12 October, CERT-FR, the alert service of France's cyber agency ANSSI, reported "numerous compromises" of the remote-control boxes in four product lines: Danfoss AK-SM and AK-SC, Copeland xweb and Dixell xweb. Its alert (we're quoting the French-language text) makes no mention of AI, and says that exposing a management interface to the internet is "contrary to good practice."

Danish manufacturer Danfoss told the trade site Cooling Post that most of the affected units were either older products or recent products running out-of-date internal software. The AK-SM800 has been end-of-life since 2021, the AK-SC355 was withdrawn in 2016, and neither gets security updates any more. On the current model, the known flaws were patched in March and August 2025, according to Infosecurity Magazine, so before Mythos. Danfoss hasn't said how the attackers got in. US firm Copeland, for its part, talks about an abandoned generation of controllers.

A supermarket fridge is a computer installed like a household appliance. You plug it in, forget it thinks, and it stays online for years with its original password.

Unplugging instead of patching

The best clue comes from the aftermath. Censys, a firm that catalogs devices visible on the internet, counted 1,825 Danfoss control panels exposed in France on 11 October. Three days later, 537 were left, according to its report as picked up by Infosecurity Magazine. Censys notes that patched versions weren't increasing: operators mostly just unplugged.

Exactly Alex Pearce's move. His own system, installed in 2018 by US firm Copeland, will need updating, he admits. When the Guardian met him, he still hadn't plugged the cable back in.

Who and why: still murky

That leaves the question of who did it. Xavier Bellekens, who runs Lupovis, a young company that sets traps for hackers, believes the attacks carry Russian hallmarks: his decoys, built with the same boxes, drew attackers coming through hosting providers known for looking the other way. But he adds straight away: "Are we certain these are Russian groups? No." No official investigation has named anyone.

Ciaran Martin, who founded the NCSC and ran it, rejects the "Wuhan moment" label. "It's a bad run, but it's not unprecedented," he told the Guardian, pointing to the WannaCry and NotPetya waves in 2017 and the ransomware surge of 2021.

That doesn't take AI off the table. The NCSC considers it "highly likely" that AI-boosted malware will target British power plants, airports and hospitals by 2028. But the expert Tal Be'ery, quoted by the Guardian, made an observation the fridges illustrate well: the surge in discovered flaws is not, so far, translating into a surge in break-ins.

The old door

So two stories have been fused into one. AI lengthens the list of known flaws every month. The fridges, meanwhile, were opened through the oldest door there is: a forgotten device left online, not updated, sometimes with the password printed in its manual.

CERT-FR warns that the campaigns against these boxes are "probably still under way." The first defense needs no AI at all: go and check what's still plugged in.

Was this article useful?

Topics covered:

SecurityAnalysis

Frequently asked questions

What happened to the supermarket fridges?
In the night of 11 to 12 October, around 700 of France's 10,000 supermarkets had their refrigeration systems thrown off remotely, according to Finance Minister Roland Lescure, quoted by AP. On 2 November, shopkeepers in the UK, Belgium, Estonia, Ireland, Finland, Luxembourg, the Netherlands and Sweden reported fridges and freezers climbing to as much as 50 °C, according to the Guardian. The same day, cold rooms belonging to the Établissement français du sang (the French blood service) were targeted at three sites, with no loss of blood products, the service says.
Did the hackers use AI?
"The key thing is that this was not an AI attack," said Paul Chichester, director of operations at the NCSC, the UK's cyber security agency, quoted by AP. The attackers used vulnerabilities that had been circulating for a very long time. Nicole Carignan of the security firm Darktrace found no evidence that AI was used, and says the techniques involved didn't require it.
Which devices were hacked?
On 12 October, CERT-FR, the alert service of France's cyber agency ANSSI, reported numerous compromises of the remote-control boxes in four product lines: Danfoss AK-SM and AK-SC, Copeland xweb and Dixell xweb. According to Danfoss, most of the affected units were either older products or recent products running out-of-date internal software. The AK-SM800 has been end-of-life since 2021 and the AK-SC355 was withdrawn in 2016.
Do we know who is behind the attacks?
No. Xavier Bellekens, who runs the security firm Lupovis, believes the attacks bear Russian hallmarks, but says he isn't certain. No official investigation has named anyone. In France, the Paris prosecutor's office has opened a case.
Are the attacks over?
CERT-FR warns that the campaigns against these boxes are probably still under way, and notes that exposing a management interface to the internet goes against good practice. After the French wave, the firm Censys saw the number of Danfoss panels exposed in France fall from 1,825 to 537 in three days, with no rise in patched versions: operators mostly just unplugged.
Julien-Pierre Noto

Julien-Pierre Noto

Entrepreneur & Voice from the Field

Julien-Pierre is an entrepreneur with over twenty years of hands-on experience in construction and real estate. For the past three years, he has been working with AI every day in an SME — not in a lab: on real cases, with real clients. Founder of ONDE AI R&D, an applied research lab on human-AI work, he publishes his methods as open source — what works and what doesn't. At Declic Media, he is the voice from the field: applied AI, the kind that has to prove its worth.

All articles by Julien-Pierre →
The free AI newsletter