The key facts
- In early November, temperatures of refrigerators and freezers were raised in dozens of European supermarkets after an attack in France in October
- Anthony Grieco (Cisco): the main threat is outdated systems no longer receiving updates, not new hacking methods
- Cisco used AI to scan 1.8 billion lines of code in eight weeks instead of roughly eight years of engineers' work
Supermarket attack as an illustration of an old problem
In early November, temperatures of refrigerators and freezers were raised in dozens of European supermarkets — this happened after a cyberattack in France in October. Anthony Grieco, director of security and trust at Cisco, commented to Euronews Next that such hacks are not new or sophisticated: they exploit existing vulnerabilities, and AI merely accelerates their discovery at the scale of many connected devices.
'Vintage' critical infrastructure
According to Grieco, smart refrigerators are only a small example of the risk; a far more dangerous version of the same problem unfolds in critical infrastructure: in hospitals' medical devices, factory assembly lines, water supply systems and digital networks. Many of these systems run on outdated hardware and software that manufacturers no longer support with security updates. 'There are many critical infrastructure systems that continue to operate but are at end of life and end of support. And the risk equation has changed dramatically over the past six months,' he noted, comparing the situation to a thirty-year-old car for which original spare parts are no longer available. To reduce risks, he urges companies to implement stricter credentials and passwords, decommission equipment that no longer receives updates, segment environments and install patches faster.
AI against AI
Grieco noted that attackers are beginning to use AI tools and agents to compromise IT systems, acting 'at a pace and speed for which humans previously needed far more time.' At the same time, he said, AI also helps defenders: Cisco works with partners, using models from companies such as OpenAI and Anthropic, to find and close vulnerabilities before criminals do. Earlier this year, Cisco reported that its teams used AI to scan 1.8 billion lines of source code in just eight weeks — a task that would have taken engineers roughly eight years. According to Grieco, AI also filters out false positives, identifying real threats with efficiency close to 100%. 'Ultimately, the bits on the network don't lie,' he concluded, emphasizing that suspicious network activity can still be detected and stopped regardless of AI's capabilities.

