While artificial intelligence represents a lever for efficiency, it also doubles as a multiplier of threats: vulnerability detection, content creation for social engineering, development of autonomous malware, exploitation of human and technical weaknesses. Not to mention shadow AI quietly growing across organizations.
SMEs and mid-sized companies are now on the front line of cyber-attacks
The cyber threat landscape is scaling up. Attacks are becoming more credible, more targeted, and harder to detect for organizations. Often less prepared and equipped, small and medium enterprises (SMEs) and mid-market firms are prime targets.
The cybersecurity of industrial environments is now at the heart of corporate concerns for operators of critical infrastructure: system convergence, interconnected networks… The boundaries between IT and OT are blurring.
The cyber-attack that struck the British automaker Jaguar Land Rover should警 us: factory stoppages for three weeks, teams on short-time, more than 5,000 suppliers and subcontractors affected, some facing bankruptcy. The financial cost is estimated in the billions of euros.
At the same time, the regulatory framework is tightening, notably with the Cyber Resilience Act, the AI Act and the NIS 2 directive, which broaden cybersecurity requirements to an increasing number of companies.
And more recently, AI agents that are now purportedly capable of executing cyber-attacks autonomously. A few weeks ago, it was suggested that AI could be so powerful as to threaten humanity itself!
Between a certain degree of fear-mongering from major tech players—who see an incentive to justify ever-increasing investments—and the real technological feats, it becomes difficult to draw a clear risk map.
AI accelerates the pace of cyber-attacks… and strengthens defense!
If AI enables more industrialized and automated attacks, its real impact on execution speed has increased, threatening organizations by rendering current response processes obsolete.
We might be tempted to see the glass as half empty: attackers will eventually gain access to the latest LLM models, giving them speed and scale never before seen.
Yet there are plenty of objective reasons to view the glass as half full from the cybersecurity perspective, because AI will also help us identify attack paths that had not been identified, better comprehend complex architectural systems, be more proactive in detection and incident response, and reduce SOC fatigue.
However, to achieve that, it is necessary to implement AI governance; identify and map data; frame how employees use it so as not to hinder innovation while reducing the risks of improper use.
Ultimately, human intelligence will always be needed to manage vulnerabilities, contextualize them within the business, and choose technological dependencies.
More than ever, the human factor will be decisive, with a need to raise awareness and train employees about the opportunities offered by AI tools, while taking into account privacy and security concerns.
Find the full Dawn Liphardt Morning briefing dedicated to Cybersecurity – AI, a new accelerator of threats
*Clément Donzel is a cybersecurity expert, advising MonAideCyber with ANSSI and President of the company OFFMYDATA
