Fridges heated to 50°C, AI renamed by decree, small businesses building without governance

Three signals this Monday. Hackers turned French supermarket fridges into ovens. Washington renames artificial intelligence by decree. And the real builders of AI are now small businesses, often without governance. I take them in order.

Signal one: fridges on fire

Overnight from Sunday to Monday, hackers pushed the temperature of connected fridges in French supermarkets up to 50°C. Food destroyed. Fires reported.

Reuters reports that several hundred stores were hit:

• Intermarché reports about 150 stores affected, with fridges held at 50°C for nearly six hours, and all affected products thrown away.
• Carrefour says 80 supermarkets were hit.
• Coopérative U confirms its stores were affected too.

The Paris prosecutor describes a large-scale attack on connected fridges at several major grocery chains. Prime Minister Sébastien Lecornu ordered ANSSI, France's national cybersecurity agency, to investigate, and says France is coordinating with its European counterparts. In the Netherlands, Ahold Delhaize, parent of Albert Heijn, reports a security issue on its own network and takes several systems offline as a precaution. At this stage no specific vulnerability has been attributed and no group has claimed the attack.

Where the breach happened

I look at the location, because it is the lesson. The breach happened in the cold chain, far from data centers and cloud providers. In refrigeration controllers, the small operational boxes that keep the milk cold.

For twenty years, a retailer's cybersecurity protected its data: customer files, payments, checkout systems. Connected operational equipment, what is called the industrial internet of things, was put on the network to be monitored remotely, often with minimal security. It still controls physical things: a temperature, a door, a pump. An attack on it steals nothing. It damages the real world.

The attacks of this era will not always steal your data. Sometimes they will simply ruin dinner.

What changes

This is exactly what the previous week's crisis exercises rehearsed. AI lab executives were privately simulating a catastrophic incident, and governments were demanding transparency. The real incident came through another door, far from any rogue model. It came from criminals who found the most exposed and least glamorous equipment in the store.

Security has become a facilities issue. A fridge issue.

For brands, food retail rests on one asset: trust. When customers learn hackers heated the fridges, their question is about whether the food is safe, long before which patch was applied. Intermarché understood immediately: throw everything out, take the loss, protect trust. It is a brand decision, made in an emergency, and it is the right one.

I keep one rule for any company running physical sites: inventory what is connected and what each piece of equipment controls. The answer to "what happens if someone takes control of it?" must exist before the incident, along with the brand decision that goes with it.

Signal two: Washington renames the technology

On October 8, President Trump posted on Truth Social that the White House considers "THE ENEMY" anyone who keeps using the term artificial intelligence instead of his preferred phrase, Super Intelligence.

I covered it on October 5. On September 29 he signed Executive Order 14434, "Inaugurating the Era of Super Intelligence," published in the Federal Register on October 2. It directs federal departments and agencies to replace artificial intelligence and AI with Super Intelligence and SI in official communications, websites, reports and policy documents.

The name changes, the definition stays

Here is the telling point. The order changes the name and keeps the definition. Super Intelligence is legally defined as exactly the same technologies artificial intelligence already covered in US law. Same systems, new label. No penalties for anyone who keeps saying AI outside government.

The transition is already messy. The White House rebranded AI.gov as "SI GOV," but the site still lives at ai.gov, and a good share of its pages still say AI.

The market is already following. Elon Musk renamed the SpaceXAI social account to SpaceXSI after a user asked him on X, and wrote that SpaceX is a superintelligence company.

The problem with the word

In AI research, superintelligence already has a precise meaning: a hypothetical system that would exceed human intelligence across a wide range of tasks. Using it as the new name for today's chatbots erases a technical distinction the field has used for decades. The name now promises more than the product delivers.

That is why this story belongs here. A name is a brand's first asset. Whoever names a technology frames the conversation about it. A government can rename a technology by decree. A company can rename itself overnight to ride the decree. But the word's meaning was borrowed from researchers who meant something else. Words have owners, even when nobody registered them.

What changes for brands

If SI becomes the official vocabulary of public contracts, tenders and regulation, every brand selling AI in the US will have to choose: follow the flag, or defend the technical term. I would advise against rushing. Adopting a word that overpromises means accepting being judged against that promise. A solid brand can use its customer's vocabulary in a tender and keep exact words in its own voice. Language is infrastructure too.

Signal three: AI's real builders are not who you think

Pax8, a marketplace serving small and mid-sized businesses, published its quarterly AI survey: 1,150 leaders of companies with 5 to 499 employees, in the United States, the United Kingdom, Germany and Sweden. The headline: small businesses are no longer just using AI, they are building it.

The numbers

• 67% of small businesses engaged with AI have built one or more solutions now in regular use.
• 19% more are prototyping.
• Only 13% stick to off-the-shelf tools.
• Among builders, 46% are building or testing agents that complete multi-step tasks on their own.
• 66% believe their business will not stay competitive without AI within three years.

Governance is not keeping up

• 35% of these companies discovered, in the past six months, AI running in their operations without anyone's approval. Shadow AI.
• Formal, documented review of internally built AI peaks at 45% in Sweden and falls to 23% in the UK.
• 49% of companies that paid for AI received at least one bill higher than expected.
• 55% of decision-makers personally spend three hours or more a week reviewing, correcting or checking AI-generated work.

What changes

The AI economy is decentralizing. The frontier is also inside a forty-person logistics company plugging an agent into its finance system. When a small business deploys an agent, it takes on work that used to belong to software vendors: maintaining what it built, securing it, and paying for the compute it consumes.

That is where the real risk accumulates. The rogue model makes headlines. The everyday risk sits in ten thousand unreviewed workflows quietly touching customer data, finance and operations. Data exposure, security and privacy combined, is the most cited risk across every market studied.

One number speaks louder to me than the rest: three hours a week of checking for more than half of decision-makers. AI saves the team time and takes some back from the leader. It is a hidden cost that appears in no return-on-investment calculation.

For a small business leader, I would propose minimum governance, doable in a week: an up-to-date list of every AI tool or agent in service, with its owner, the data it accesses and what it may do on its own. One line per tool. It is little, and it is already more than most do.

The future of AI governance will be written in small businesses, in agent logs and spreadsheets, far more than in executive orders.

The thread of the day

An attack that heats fridges instead of stealing files. A name imposed by decree on an unchanged definition. Small businesses building faster than they can control.

The common thread: the gap between what is visible and what matters. Risk lives in the forgotten equipment, the borrowed word, the tool plugged in without review. That is where to look.

The noise filtered. The signal kept.