In the wake of the cyber attack impacting hundreds of stores at French supermarket chains including Carrefour, Intermarché, Coopérative U and Netto, causing refrigerators to defrost millions of euros worth of food, the companies are now working to identify fridges that are vulnerable to cyber attacks.

The services that we all rely on are increasingly a prime target for cyber attacks and they cause widespread disruption. The freezers in around 500 French supermarkets did exactly what they were told, a command arrived (defrost, and keep defrosting) and the equipment executed it for six hours, at temperature up to 80°C.

Before any command touches a physical asset, two checks have to happen, right where it lands. The first is to ask whether the command is genuine. Did it come from an authorised origin, unaltered, not replayed, not injected by a compromised system along the way? A network connection or a login on an intermediate server proves none of these points. The command itself must carry the proof, and the asset must verify it.

The second check is whether the command is allowed. Even a genuine command can be wrong; a stolen credential, a careless operator, a faulty AI writeback. A freezer has physics and the defrost command lasts minutes, frozen setpoints, therefore, need to stay below zero. Anything outside that envelope should either be refused or require a second approval. If either check is failed, the command should never execute.

On Sunday night, neither of these checks existed where it mattered. So, a few lines of instruction became millions of euros of food in the bin.

Zero trust in operational technology means not trusting the network, not trusting the server, but verifying every command at the asset, and checking it against policy, before it runs.