Search

The Fragility of a
The Fragility of a

Thousands of cancelled flights, disruptions to banking services, inaccessible communication…

Designing the right to
Designing the right to

Imagine your banking app classifies a meal with friends as a business expense. Or an artificial…

When football decides
When football decides

A few days ago, I tried to access a client’s website from my mobile phone. The browser appeared to…

The Fragility of a Hyperconnected World

The Fragility of a Hyperconnected World

Thousands of cancelled flights, disruptions to banking services, inaccessible communication platforms and malicious code distributed through the websites of major organisations. The events of recent days have once again highlighted just how much we depend on the proper functioning of our digital infrastructure.

Between 13 and 17 November, organisations such as Microsoft, United Airlines, JPMorgan Chase and the US Federal Aviation Administration (FAA) experienced significant technological disruptions. Although their temporal proximity fuelled speculation about a possible coordinated cyberattack, US authorities pointed to technical issues and found no evidence to support that hypothesis.

At the same time, on 16 November, a vulnerability in the Optimizely web experimentation platform made it possible to distribute malicious JavaScript code through the websites of at least 178 organisations. In this case, we are indeed dealing with a confirmed attack on the digital supply chain.

The distinction matters. Not every technological disruption is the result of a cyberattack, nor does the temporal coincidence of several incidents mean that they are related. Confusing facts, hypotheses and speculation makes it harder to understand the problem and may encourage the spread of misinformation.

However, beyond the origin of each incident, there is a reality that deserves our attention: both a deliberate attack and a technical failure can have consequences that extend far beyond the boundaries of the organisation where they originate.

And this is precisely where cybersecurity ceases to be an exclusively technical matter and becomes a systemic challenge.

Being connected is not the same as being independent

A company may have its own systems, apply good security practices and have a highly qualified technology team. Yet its day-to-day operations depend on an extensive network of external services: cloud providers, authentication systems, payment platforms, content delivery networks (CDNs) and third-party tools.

This interdependence is not necessarily negative. It has allowed us to develop products more quickly, reduce costs and access technological capabilities that we would struggle to take on individually. The problem arises when this interdependence coexists with a high degree of technological concentration.

Two companies operating in entirely different sectors may use the same hosting provider, depend on the same authentication system or share a communications infrastructure without being aware of it.

The European Central Bank itself warns of this risk in the financial sector: the concentration of technology services among a small number of providers can mean that a disruption affects numerous entities simultaneously.

Moreover, using different providers does not necessarily guarantee greater independence. They may, in turn, rely on the same infrastructure, technologies or external services.

We are therefore faced with a paradox: the more organisations we digitalise, the more we do not necessarily build independent infrastructure, but rather expand our dependence on existing infrastructure.

The question is no longer merely who we depend on, but how many organisations depend on the same things we do.

From individual risk to systemic risk

Systemic risk emerges when a localised problem has the capacity to spread and compromise the functioning of an entire system.

Imagine that an authentication service provider suffers an outage. Its customers lose access to certain applications, which brings internal processes to a standstill, delays commercial operations and affects other services that rely on them. The initial incident triggers a succession of failures that reaches even organisations with no direct relationship with the affected provider.

This is what we know as cascading failures.

A particularly illustrative example was the CrowdStrike incident in July 2024. A faulty update affected 8.5 million Windows devices, less than 1% of the global total. Yet its impact reached airlines, hospitals, banking institutions and other essential services.

What was significant about that incident was not only the number of devices affected, but also the importance of the functions they performed within the system.

From this perspective, traditional risk assessment models are insufficient if they are limited to analysing the likelihood of an incident occurring and its direct impact on an organisation. It is also necessary to consider its capacity to spread, its indirect consequences and the mechanisms in place to contain them.

A society may consist of individually secure systems and still be collectively vulnerable.

Understanding systemic risk therefore requires us to stop assessing the security of each component in isolation and begin analysing the behaviour of the whole when one of them fails.

Artificial intelligence changes the speed of risk

Artificial intelligence is not only transforming the nature of cyberattacks, but also the speed at which they can be carried out.

Identifying vulnerabilities, generating malicious code or preparing social engineering campaigns are tasks that can be automated and carried out at a scale that previously required far greater resources.

The problem lies not only in attackers having more advanced tools, but in the growing asymmetry between the speed at which an attack can be executed and organisations’ capacity to respond.

While certain offensive actions can be automated, a company’s response may still depend on validation processes, internal protocols, authorisations and human decisions.

The report published by Google Threat Intelligence Group and Mandiant on 16 November confirms this development. Researchers have identified malicious actors using multi-agent systems to automate different phases of an attack, from identifying vulnerabilities to exploiting them without human intervention.

In an interdependent digital ecosystem, this acceleration takes on an additional dimension: it reduces the time available to contain an incident before its consequences reach other organisations.

Paradoxically, artificial intelligence can also help reduce this asymmetry through anomaly detection, early threat identification and the automation of certain defensive responses.

The challenge is to incorporate these capabilities without giving up human oversight or introducing new vulnerabilities resulting from automation itself.

Digital security no longer depends solely on the ability to prevent an incident, but also on the ability to react before it spreads. And that window of time is becoming ever smaller.

The paradox of digital efficiency

Digital transformation has been driven, to a large extent, by the pursuit of efficiency: reducing costs, centralising operations, automating processes and eliminating duplication.

These decisions make sense from an economic and operational perspective. However, when efficiency becomes the main design criterion, we may end up building systems that are less prepared to deal with exceptional circumstances.

Consider an organisation that centralises all its processes on a single platform. It reduces management complexity, makes maintenance easier and improves productivity. But what happens when that platform is no longer available?

From an efficiency perspective, having alternative systems, additional capacity or manual procedures may seem unnecessary. From a resilience perspective, these resources represent mechanisms that make it possible to continue operating when normal conditions no longer apply.

What appears redundant under normal conditions may prove indispensable in a crisis.

This does not mean that we should abandon optimisation or indiscriminately multiply infrastructure. It means that we must incorporate recovery capacity as another criterion in technological decisions, alongside cost, performance and scalability.

The real challenge lies in finding a balance between efficiency and resilience. Because a system optimised to operate under ideal conditions is not necessarily a system prepared to operate in the real world.

From cybersecurity to collective resilience

If digital risks extend beyond the boundaries of organisations, the strategies for addressing them should do so too.

Collective resilience means understanding that an organisation’s ability to prevent, deal with and recover from an incident depends, in part, on the decisions made by other actors in the digital ecosystem.

This perspective requires responsibilities to be assumed at different levels:

  • Businesses must understand their technological dependencies, assess the risks associated with third parties and have continuity plans that are regularly tested.
  • Technology providers must take responsibility for their position within the ecosystem, incorporating fault-isolation mechanisms, transparency during incidents and recovery procedures.
  • Public institutions must promote frameworks for cooperation, the exchange of threat information and coordinated response mechanisms across sectors and countries.

In Europe, initiatives such as the NIS2 Directive and the DORA Regulation reflect this shift in perspective by incorporating third-party risk management, incident reporting and operational resilience within their respective scopes of application.

However, regulatory compliance is only one part of the response. We also need a culture of collaboration in which sharing information about incidents, vulnerabilities and lessons learned helps improve the security of the whole.

Collective resilience is not about ensuring that no system ever fails, but about preventing the failure of one from compromising the operation of all.

Conclusion

The events of recent days do not call into question the benefits of digitalisation, but they do invite us to reconsider some of the criteria by which we assess technological progress.

We have learned to build systems that are faster, more efficient and more connected. However, the ability to maintain our operations in the face of adverse circumstances should also be an indicator of digital maturity.

This requires us to broaden our conception of cybersecurity, incorporate resilience from the design stage, and acknowledge that absolute technological independence is virtually impossible in an interconnected world.

Perhaps we should begin to measure digital progress not only by the possibilities technology offers us when everything works, but also by our ability to keep functioning when something fails.

Become a member

Get the latest updates straight to your inbox. No spam.

What is the point of building a society that is increasingly technologically advanced if, at the same time, we make it increasingly vulnerable?

Comments
Comment