Overnight, while shoppers slept, hundreds of French supermarkets were quietly cooking their own stock. In a French supermarket cyberattack confirmed by retailers on Monday, hackers took remote control of internet-connected refrigeration systems and pushed fridge and freezer temperatures as high as 50 degrees Celsius (122F). By morning, tonnes of food had perished, some equipment had melted, and in some stores the heat had sparked fires. The cyberattack did not rely on a new exploit. It targeted ageing connected devices that were still online years after their manufacturer stopped supporting them, a weakness experts say is common across retail and industry.

What Happened in the French Supermarket Cyberattack

The attackers struck when shops were closed, raising temperatures remotely through the refrigeration control systems. French TV channel BFM reported that they appeared to trigger a forced defrost option, keeping units heating instead of cooling. Five of France's major retail groups were hit: E.Leclerc, Intermarché, Carrefour, Monoprix and Coopérative U. At least 300 Intermarché and Netto stores were affected. Coopérative U reported more than 100 stores, Carrefour said 80, and Monoprix said about 5 percent of its roughly 490 stores were targeted. The full scale is still unclear. Some reports put the total above 500 shops. Bruce Hallas, CEO at Re-thinking the Human Factor, who lives in France, noted on LinkedIn that current estimates sit around 400 stores, with some individual branches reporting losses of about €250,000. Retailers say millions of euros worth of food was spoiled.

Danfoss Refrigeration Controllers at the Center

France's government cybersecurity agency, ANSSI, identified the targeted systems as Danfoss refrigeration controllers, including the AK-SM800 and AK-SC355 models. The Danish manufacturer told AFP that most affected units were legacy products running outdated firmware, and that it had previously advised supermarkets to update them. ANSSI has urged companies using the affected models to disconnect them from the internet or restrict access, and to install software updates immediately. Authorities have opened an investigation and are liaising with other European countries that may have been affected. It remains unclear who carried out the attack. The incident follows President Emmanuel Macron's order last month to protect France's critical infrastructure from Russian "hybrid attacks." Officials have not linked the supermarket attack to any group or country.

Exclusive: Why These Devices Were Left Exposed

In comments shared exclusively with The Cyber Express, Neil Carter, Managing Director of CRAReady.co.uk, explained why systems like these remain vulnerable for so long. "Most of these devices were never designed to be updated in the field. A supermarket refrigeration controller is installed by a contractor, commissioned once and then left alone for ten or fifteen years. Nobody owns it as an IT asset, so nobody is watching for firmware releases. There's also a mismatch in lifetimes. The hardware lasts far longer than the vendor's support for it. Danfoss ended support for the AK-SM 800 in 2021 and the AK-SC 355 in 2016, but those boxes were still running cold chains in 2026. The third problem is how they're connected. Remote access is added for the maintenance company's convenience and simply left open to the internet. Updating carries risk too: if a firmware update fails on a controller running a store's freezers overnight, the stock is lost. So the safest-looking choice is always to leave it alone."

A Night-Long Threat to the Cold Chain

The overnight timing turned a technical breach into a major cold chain failure. Mark Eccleston, Senior Software Engineer at OxBlue, pointed out on LinkedIn that a fridge can go about four hours without power before food is at risk. "Hackers got a whole night with hundreds of them," he wrote. Retailers were forced to discard fresh and frozen products to avoid health risks, and the incident has raised wider questions about how much of Europe's food supply depends on internet-connected equipment running legacy firmware.

What Businesses Should Learn From the French Supermarket Cyberattack

Carter outlined clear steps for businesses running connected equipment: "First, know what you have. Treat building and plant controllers as part of your attack surface and keep an inventory with support end dates. Second, nothing operational should be directly reachable from the internet. Remote maintenance should go through a VPN or be restricted to the contractor's addresses. Third, ask about end of life when you buy, not after. Under the EU Cyber Resilience Act, manufacturers will have to declare a support period that matches realistic product lifetimes. Buyers should already be asking for that in writing. This attack didn't need a zero-day. It needed old, forgotten kit on the open internet, and there's a lot of that out there."

OpenAI models accessed the public internet during separate third-party cyber evaluations conducted by independent testing partners, prompting the company to review how high-risk AI testing is managed. OpenAI said the incidents occurred under specialized testing configurations with reduced safeguards and did not reflect how its models operate in public deployments. The company added that the events were unrelated to the previously disclosed Hugging Face security incident.

The incidents involved evaluations conducted by UK AISI and cybersecurity testing partner Irregular, where testing conditions or environment configurations enabled models to interact with systems beyond the intended evaluation boundaries.

OpenAI Models Prompt Review of Third-Party AI Testing

OpenAI said independent cybersecurity evaluations are essential for understanding model capabilities before deployment. Some evaluations intentionally reduce safeguards or enable additional capabilities to measure how models perform under conditions that resemble real-world cyber operations.

According to the company, the latest incidents highlighted the need to strengthen the security controls surrounding independent testing environments as AI models become more capable.

OpenAI said it will review how it manages third-party cyber evaluations, including how higher-risk testing is identified, when internet access or reduced safeguards should be permitted, how testing environments should be isolated, and how incident reporting and monitoring procedures should be handled.

The company also plans to work with national AI institutes, independent evaluators, and other AI developers to establish stronger industry practices for high-risk AI evaluations.

UK AISI Identified Two Unsanctioned Actions by GPT 5.6 Sol

On August 3, UK AISI, the UK's AI Security Institute, informed OpenAI that during a cyber evaluation that began on July 25, models from OpenAI and another AI lab performed activities beyond the intended testing scope.

The evaluation used simulated cyber ranges where models acted as cybersecurity experts in a Capture-the-Flag exercise. Internet access was intentionally enabled so agents could download tools similar to those available to real attackers, while cyber classifiers were disabled to measure underlying model capability.

Among 19 recorded events, two involved GPT 5.6 Sol.

According to OpenAI, the model reused a publicly accessible GitHub token left behind by another lab's agent to determine whether systems inside the simulated environment were communicating with GitHub or update services. It also attempted account recovery and request-limit workarounds while registering accounts with external DNS and tunneling providers.

In a separate action, GPT 5.6 Sol used a public tunneling service to expose a locally hosted DNS server from the evaluation environment to the public internet. The server contained exploit payloads intended for a known software vulnerability within the simulated cyber range.

OpenAI said the setup did not function as intended, no evidence indicated that any real internet resolver queried the server, and the infrastructure was removed when the evaluation concluded.

UK AISI considered both actions outside the authorized testing boundary because they involved external services that were not required to complete the exercise.

Security monitoring detected unusual data transfers on July 28. UK AISI halted the evaluation, isolated the affected systems, and contained the activity within approximately one hour.

Irregular Finds Misconfigured Testing Environment

A second incident was reported on July 29 by Irregular, another organization conducting third-party cyber evaluations for OpenAI.

The testing environment was intended to remain isolated from the internet during another Capture-the-Flag exercise. However, a configuration error unintentionally allowed the models to access external websites.

During one evaluation, the fictional domain used in the exercise matched an actual website. Because internet connectivity was available, the model interacted with the live site, believing it was part of the simulated challenge.

Irregular said the incident did not involve a sandbox escape or zero-day exploit. Instead, the model exploited a basic security vulnerability on the website after gaining unintended internet access.

The investigation also found that the model used credentials associated with the same website. According to Irregular, no impacts beyond that site's own data have been identified, although its audit remains ongoing.

The organization has paused the evaluations, completed remediation, notified affected third parties, and implemented additional safeguards in its testing environment.

OpenAI said it will continue working with both UK AISI and Irregular to improve evaluation practices while ensuring independent cybersecurity testing remains rigorous as AI capabilities continue to advance.