"Cybersecurity operations today involve vast amounts of real-time data," explains Pierre Girard. "Generally, in a SOC (Security Operations Center), we monitor a lot of things: network events, server logs, endpoint data, firewall alerts, etc. It’s like a cockpit for cyber defense." Managing such data volumes manually is impractical. Here, AI becomes essential. Machine learning enhances detection by recognizing not just known patterns, but "things similar to" known threats. "That's the power of AI," says Pierre Girard. "It doesn’t just react, it anticipates."

Yet AI is a double-edged sword. "It empowers attackers too," Pierre Girard warns. From deepfakes to automatic malware generation using large language models (LLMs), attackers are now armed with tools that can produce endless variants of malicious code, automating and industrializing cyberattacks. The barrier to entry is lowered: "You don’t need to be a coder anymore to orchestrate complex attacks."

From reactive to proactive cybersecurity

While AI enhances detection, its true value lies in shifting cybersecurity from reactive to proactive. Alexis Valero highlights this transformation: "Our use of AI at rzilient moves us beyond patching and troubleshooting. We now operate in a logic of continuous security audits and predictive maintenance." Their system can anticipate hardware failures, monitor component degradation, and recommend pre-emptive action; enhancing both security and performance.

"The beauty of AI is in orchestration," Alexis Valero continues. "We aggregate data from multiple tools to detect anomalies, recommend actions, and even automate responses. This not only reduces ecological impact by preventing hardware waste but also mitigates the risk of cyberattacks tied to system failures."

Pierre Girard echoes this, noting how minimizing attack surfaces can serve both cybersecurity and Green IT objectives. "Removing unnecessary services from operating systems not only reduces vulnerabilities but also cuts energy use. It's a win-win." He cites another well-known cybersecurity practice: eliminating zombie IT, unused apps or machines that consume resources without purpose.

He adds: "At Thales, we emphasize reducing the digital footprint of IT systems. We support our clients on minimizing superfluous processes and trimming down to core necessities. It’s a strategy that supports both cybersecurity needs and Green IT ambitions. This also includes keeping systems updated to avoid potential vulnerabilities."

Alexis Valero elaborates: "There’s a real need to build a culture of anticipation. Through constant auditing and proactive systems, we can prevent breakdowns before they occur. We tell our clients that it’s not just about cost-saving, this is about resilience and responsibility."

Scaling sustainability with intelligent choices

The conversation inevitably turns to the environmental cost of AI itself. Pierre Girard is straightforward: "LLMs are extremely energy-intensive. Efficiency is one pillar of Green IT, making things run faster, on smaller models. But the other is sufficiency: doing only what’s truly needed." According to him, the obsession with integrating LLMs into everything is misplaced. "Traditional AI, especially classic machine learning, often suffices for pattern detection tasks."

At Thales, sustainable and responsible AI is more than a concept; it’s a strategic position. The company participated in the drafting of the Afnor Spec on Frugal AI and contributes to the Coalition for Sustainable AI launched at the AI Action Summit in Paris in 2025. "We advocate exiting the LLM hype when it’s unnecessary," Pierre Girard emphasizes. "In cybersecurity, AI (not Generative AI) is excellent at pattern recognition, so let’s use the right tool for the job."

He details: "AI 'à l’ancienne', as we say, is highly effective in many detection scenarios and much more energy-efficient. There’s a growing tendency to add AI where it’s not needed. At Thales, we actively promote thoughtful use of AI, avoiding unnecessary overhead and opting for scalable, frugal solutions."

He further clarifies: "People need to be educated on this. There’s a tendency to think 'more data, more AI' means better results. But in cybersecurity, quality matters more than quantity. A smarter, leaner system is often the safest and most sustainable one."

Alexis Valero supports this perspective, acknowledging the tensions between performance and sustainability. "Green IT demands lightweight, low-power systems. But AI requires powerful hardware and data processing. It’s a paradox. We believe Europe has an opportunity here: to push for frugality in infrastructure and design security into systems from the start."

AI's double role: amplifier and equalizer

Both experts agree that while AI increases cyber risk by amplifying the power of attackers, it can also act as a great equalizer for defenders. "The scale and automation it enables are unmatched," says Pierre Girard. He points to AI's ability to detect patterns that humans would miss, which can lead to early warnings and predictive action.

Alexis Valero underscores the importance of this capacity: "Thanks to AI, our detection range is vastly expanded. We are now able to catch issues, on computer for examples, months before they become critical. Security becomes a continuous, intelligent process rather than an occasional check-up."

He highlights how performance degradation can be a risk: "When a component like RAM begins to fail, performance drops, increasing the risk of errors and potential security breaches. Predictive AI allows us to intervene before a failure occurs." He concludes: "By orchestrating incident simulations and strengthening active monitoring, AI lets us automate detection, remediation, and even user education. It's not just about alerting, it's about solving."

Pierre Girard agrees with Alexis Valero. He points out that with LLMs, attackers can now generate millions of slightly varied versions of malware, making detection harder, "But defenders can also harness AI to cluster and identify these variants. The same power that aids attackers can - and must - be used by defenders, too." 

And to conclude: "The issue," he adds, "is often human error. Social engineering attacks, like phishing or deepfakes, exploit psychological weaknesses. AI can intervene earlier in the process, flagging odd behavior or language and helping teams act fast."

Toward a frugal and resilient digital future

As AI continues to transform cybersecurity, the need for thoughtful integration becomes urgent. "If we had simpler, more frugal systems to begin with, we’d have fewer cybersecurity issues," Alexis Valero observes. More complex systems create more potential entry points. "That’s why rethinking IT architecture is also a cybersecurity strategy."

For Alexis Valero, the challenge also lies in finding the right partners to connect their solution to: “To reduce the environmental footprint of our services based on AI, we are actively seeking infrastructure options that are, in the end, built on greener foundations.”

Thus, the challenge in making cybersecurity more sustainable is not to eliminate AI, but rather to find the right balance at every level of the information system. A more frugal system will be simpler to secure and will require less intensive AI integration. Organizations must also make deliberate choices not to default to highly energy-consuming AI solutions powered by LLMs. Alternative paths exist, including smaller models (SLMs) that focus on specific tasks..

In the end, sustainable cybersecurity is less about flashy innovations than it is about balance: between anticipation and restraint, automation and intention, speed and frugality. With AI as both a lever and a risk, the coming years will test whether organizations can build systems that are not just smarter, but resilient.

‍

Pierre Girard is digital eco-design director at Thales Cyber & Digital.

Alexis Valero is co-founder and CEO of rzilient.

Pierre Girard and Alexis Valero will be speakers at GreenTech Forum Brussels 2025 on the round table entitled “Cybersecurity: what compatibility with Green IT in the era of AI?”.