AI in cybersecurity: faster threats and defences for SMEs

Analysed source: www.silicon.fr. For a leader, the value lies in the decision the analysis helps clarify.

In the Silicon.fr opinion piece “Cybersecurity: how AI accelerates threats… and helps cyberdefenders”, published on 22 September 2026, Clément Donzel describes AI's expansion of attack vectors and defensive options. For industrial SMEs and mid-market companies, the signal calls for a review of AI governance, data mapping and response processes. Leadership must decide how to integrate controlled AI into the security operations centre (SOC), as high-speed automated attacks may outpace existing responses.

AI as a threat multiplier for industrial SMEs and mid-market companies

AI models can identify vulnerabilities, generate social-engineering content and design autonomous malware. Silicon.fr notes how malicious actors exploit these capabilities for more credible, targeted attacks. SMEs with fewer resources than large groups become attractive targets, especially where IT and operational technology converge. Internal shadow AI also increases exposure without teams realising it. explore governance and risk.

Execution speed: why response processes become outdated

AI shortens the attack cycle from vulnerability identification to exploitation to minutes. Traditional manual response steps struggle with this speed: an incident may spread before the SOC records the anomaly. Inaction risks losing control over critical systems. The article cites a three-week shutdown of Jaguar Land Rover factories, with thousands of employees placed on temporary layoff and thousands of suppliers at risk. explore governance and risk.

AI for proactive detection and reduced SOC fatigue

The technology used by attackers can also reveal attack paths invisible to human analysts. Correlation and supervised-learning algorithms can continuously scan complex architectures, anticipate intrusion scenarios and trigger early alerts. Automating triage and prioritisation reduces analysts' information overload and decision fatigue, improving response quality.

Governance and data control: essential prerequisites

Clear AI governance must come first: map datasets, define who can manipulate them and enforce strict access controls. Without this framework, innovation may be constrained or create vulnerabilities. Compliance with the Cyber Resilience Act, AI Act and NIS 2 Directive becomes a resilience lever.

Hypothetical scenario: a production line under threat

Imagine an industrial SME running an automated assembly line. A malicious actor uses an LLM to generate a script disabling safety controls in programmable controllers, injected through a misconfigured VPN. Without AI detection, the incident remains unnoticed for hours, stopping production. In the hypothetical scenario, an AI module analysing network flows and controller behaviour detects, isolates and corrects the anomaly before operational impact.

Practical decisions to implement

• Establish AI governance covering data and access-right mapping.
• Deploy AI event correlation within the SOC to automate alert prioritisation.
• Train operational teams on generative-model risks and cybersecurity practices.
• Audit remote access points (VPN, RDP) and use AI-based controls to identify abnormal behaviour.
• Include Cyber Resilience Act and NIS 2 requirements in the digital-transformation plan.