Written by: Maarten Botterman

One reality is becoming impossible to ignore: the pace of AI-driven change is outstripping the speed at which our resilience mechanisms were designed to operate.

For decades, Internet resilience discussions focused primarily on redundancy, uptime, and cybersecurity. Those remain essential. But the emergence of generative and increasingly agentic AI systems changes the scale, speed, and systemic nature of the challenge. Resilience can no longer be treated as a secondary technical consideration. It must become a foundational design principle for digital society itself.

Recent developments demonstrate why. Reports on AI-enabled cyber operations, including Claude-assisted exploitation research and initiatives such as Project Glassdoor, illustrate how AI systems are accelerating vulnerability discovery, automating attack chaining, and enabling adversaries to identify weaknesses across interconnected ecosystems at unprecedented speed. The assumption that human-paced review and response mechanisms are sufficient is rapidly becoming outdated.

We are entering an environment where autonomous or semi-autonomous systems can identify, test, and exploit vulnerabilities faster than many organizations can patch or coordinate mitigation. Open-source components as well as widely deployed software stacks are particularly exposed. This creates a growing asymmetry between attacker capability and existing defensive capacity with implications not only for enterprises and governments, but for the stability of the Internet itself.

At the same time, AI is also becoming indispensable for defense. AI-driven detection and mitigation systems are already improving incident response, anomaly detection, and operational monitoring. Yet defensive AI introduces new risks of its own: bias, explainability gaps, false positives, model manipulation, and overdependence on opaque automation. As a result, resilience increasingly depends not simply on deploying AI, but on integrating it responsibly into broader operational, governance, and trust frameworks. Up and beyond continuous evolving to the robustness of our digital environments in itself.

This is precisely why the work of the Marconi Society Internet Resilience Institute matters.

The Institute’s recent report, Building Global Resilience: Strengthening the Internet’s Capacity to Withstand Disruption, highlights that resilience failures rarely stem from a single technical fault. Instead, they emerge from hidden interdependencies spanning cloud infrastructure, routing, energy systems, logistics, governance, and human operations.

AI dramatically intensifies these interdependencies.

Training and operating advanced AI systems requires enormous compute capacity, data center expansion, stable electricity supply, high-capacity connectivity, cloud concentration, and increasingly complex supply chains. The report already notes how AI-related compute demand is reshaping assumptions about energy planning and digital infrastructure dependency. In practice, this means resilience can no longer be viewed purely within the boundaries of “the Internet sector.” Energy resilience, cloud resilience, financial resilience, and Internet resilience are now deeply intertwined.

This also changes how we should think about preparedness.

The Marconi Society’s “Life of a Packet” mapping exercise demonstrates how even a simple online interaction depends on multiple hidden layers of infrastructure and governance. In an AI-enabled world, those dependencies multiply further: model providers, inference infrastructure, training pipelines, orchestration systems, data integrity mechanisms, and autonomous machine-to-machine interactions become part of the dependency chain.

The implication is clear: prevention alone is no longer enough.

Future resilience strategies must emphasize rapid recovery, graceful degradation, operational adaptability, and cross-sector coordination. The question is not whether disruptions will occur, but whether societies can continue functioning while systems are degraded, attacked, or temporarily unavailable.

This requires a shift in mindset:

• from isolated cybersecurity toward systemic resilience;
• from static compliance toward continuous measuring, testing and rehearsal;
• from organizational silos toward cross-sector collaboration;
• and from reactive incident response toward anticipatory risk management.

The WSIS+20 process and the Global Digital Compact increasingly recognize resilience as a prerequisite for digital inclusion, trustworthy AI, and sustainable digital transformation. The Marconi Society’s WSIS 2026 session taking place in Geneva, Switzerland, correctly identifies resilience as the operational layer underpinning all digital policy ambitions. Digital economies, AI deployment, e-government, and critical infrastructure all depend on a resilient Internet infrastructure.

The future of resilience therefore lies not in attempting to eliminate all failures. That is unrealistic. Systems should be built in the most robust possible ways, while keeping a keen tab on measuring what’s happening within the infrastructure to detect anomalies as soon as it happens. Next, it lies in building systems – technical, institutional, and societal – that can absorb disruption, recover rapidly, learn continuously, and maintain justified trust under stress.

This is ultimately a global challenge. The Internet remains a shared infrastructure for humanity: One World, One Internet. Ensuring that it remains stable, secure, and dependable in the age of AI will require cooperation across governments, industry, the technical community, academia, and civil society.

The pace of risk is accelerating. Our resilience efforts must accelerate faster.