Page language: Italiano English Français Deutsch
The pointAI and hacking: why the connected home is the easiest target, and how to make it hard to attack
By the FAIND newsroom ·
In recent weeks someone has heated the refrigerated cabinets of supermarkets across Europe. It took no genius: factory passwords and known flaws. The interesting part is something else. What takes time and skill today will take far less with artificial intelligence, and the connected home is full of devices with the same weaknesses. Home automation is not the problem. Home automation left as it came out of the box is.
In brief
- The attacks on supermarket fridges were an accidental test: industrial devices connected to the internet, never updated, with factory passwords.
- With AI, finding and exploiting these weaknesses gets cheaper: faster scans, more convincing scams, attacks on many devices in parallel.
- The easiest flaws are still the most common ones: factory passwords, devices not updated, remote access left open.
- The defences that work are few and cheap: different passwords, a guest network, automatic updates, remote access off when you don't need it.
- Home automation stays useful and safe if you treat it like a home installation: set it up once, properly, and check it now and then.
What happened to the fridges, in two lines
From mid-October, and again overnight between 1 and 2 November, the cold-chain control systems of supermarkets in several European countries were tampered with remotely. In one case analysed by Ransom-ISAC researchers, two administrator logins changed 743 settings on 31 devices: temperatures set to 110 and 150 degrees, alarms muted, keypads locked. Access was possible with factory credentials that had never been changed. Other controllers of the same type had a flaw that had been known for months. Who is behind it has not been established.
Why AI changes the picture
An attack like this has always had three costs: finding exposed devices, working out how to get in, repeating the operation at scale. AI lowers all three.
Finding. Models can read manuals, security advisories and code, and match a device model to an already published flaw. Getting in. Writing the code that exploits a known flaw, once specialist work, is a task coding assistants keep getting better at. Repeating. A program that tries thousands of devices works all night without tiring.
Then there is the part that concerns people, not machines: scam emails and messages in flawless English, voices cloned from a few seconds of audio, convincing fake tech support. An attack on a home can start from the phone of someone who lives there.
Why the connected home is exposed
Not because home automation is fragile, but because of how it is used. A home setup often has dozens of devices from different brands: bulbs, plugs, cameras, thermostats, fridges, TVs, voice assistants. Each has its own software and its own password. Many are installed once and forgotten. Some are cheap models whose updates stop after a couple of years.
The real risk is not someone melting your ice cream. It is the forgotten device becoming the way in to the computers and phones on the same network, or one piece of a network of devices used by others. It has been happening for a while: in 2016 the Mirai network took control of hundreds of thousands of weak devices, cameras and routers included.
The fixes that work, in order of effectiveness
1. A different password for every device. Long, unique, never the factory one. A password manager remembers them for you.
2. A guest network for home automation. Almost every router offers one. Bulbs, fridge and cameras live there, apart from computers and phones: if one gives way, the damage stays outside the important part of the home.
3. Automatic updates on. Known flaws get fixed by the manufacturer: the device that isn't updated is the one AI finds first.
4. Remote access off if you don't need it. If you don't control the fridge from outside the house, switch it off, and the same goes for UPnP on the router. What can't be reached from the internet can't be attacked from the internet.
5. Check the support period before buying. Ask for how many years the maker guarantees security updates, and prefer known brands to anonymous products. In Europe the Cyber Resilience Act, in force since 11 September 2026 for reporting, requires manufacturers to report exploited flaws within 24 hours.
6. A check twice a year. Which devices are still connected? Which are no longer used? Unplug those.
Home automation is still a good idea
Controlling lights and heating, noticing a water leak, knowing the fridge door is open: these are real benefits. The rule is the same as for any installation: you design it once, secure the critical parts and know what is in it. Anyone who does is already far harder to attack than average, and for an attacker, AI or not, the easiest target always wins.
Frequently asked questions
Does artificial intelligence make hackers more dangerous?
It makes cheaper what they already did: searching for exposed devices, writing code to exploit known flaws, creating credible scam messages. The basic defences, unique passwords and updates, stay effective.
Can my smart fridge be hacked?
In principle yes, like any connected device with a factory password or no updates. The attacks of October and November hit industrial controllers in supermarkets, not home fridges.
How do I protect my home automation?
Change factory passwords, put devices on a guest network on the router, turn on automatic updates and switch off remote access if you don't need it.
Should I give up connected appliances?
No: just set them up properly. The risk depends mostly on how they are configured and maintained, not on being connected.
Sources
- Help Net Security: Hackers abuse refrigeration controllers to send spam and route proxy traffic
- Adnkronos: Attacchi hacker ai frigoriferi, a rischio anche quelli di casa: 5 cose da fare per evitare intrusioni
- The Irish Times: After the SuperValu cyberattack, might your smart fridge be the next target?
- Euronews: UK likely to face AI-powered cyberattacks on critical infrastructure by 2028, cyber chief warns
Previous editions
- Revolut and GLM: what actually happened, and what the headlines got wrong 11 November 2026Revolut says about 2,000 engineers now default to a self-hosted GLM model. The facts, five claims going around that don't hold up, and what it means.

