How do you design for resilience in the era of machine decision-making?

Critical industries are supported by an invisible architecture – energy, finance, and logistics now depend on AI systems that make decisions faster than humans can react. And building resilience into that architecture is becoming one of the tech industry’s most urgent challenges.

The team over at Black Hat MEA recently interviewed Nikk Gilbert (CISO at RWE). He described a growing risk in digital systems that rely on automation and artificial intelligence:

“The risk that keeps me up at night is trust in machine decision-making. We’re handing over authority to AI systems in finance, logistics, and energy faster than we can test the edges.” 

And because of that, he added, “rather than bias or privacy, the real danger is what happens when these systems act on poisoned or manipulated data at machine speed.” 

AI systems already direct logistics, financial flows, and energy grids. They can respond in milliseconds and execute thousands of decisions before a human notices a problem. One compromised data stream can ripple through entire networks before intervention is possible. So designing systems that maintain continuity – the ability to withstand disruption and keep critical services operating – has become essential.

The cost of failure keeps climbing

The 2025 annual outage analysis from the Uptime Institute shows that although large outages occur less frequently than in past years, each incident carries higher financial impact. 

More than half of organisations surveyed reported that their most recent significant outage cost over USD $100,000. One in five faced costs above $1 million. Complex, interconnected environments mean that when things go wrong, the damage spreads fast. 

Physical infrastructure adds another layer of fragility. About 99% of intercontinental data travels through subsea cables. In early 2025, damage to cables in the Red Sea disrupted global internet traffic. Operators rerouted data, increasing latency and reducing quality of service. Recorded Future’s Insikt Group documented 44 public reports of cable damage in 2024-2025; caused by anchor dragging, severe weather, and regional conflict.

In that conversation with Black Hat MEA, Gilbert emphasised that resilience requires more than technical compliance:

“Compliance is not the same as security. Legally secure means you passed the audit. Actually, secure means you can take a hit and keep going. You can be fully compliant and still one mistake away from catastrophe…the real test is resilience, not paperwork.” 

A way to see weak spots 

The Cyber Defense Matrix, created by Sounil Yu, offers a clear way to visualise security posture. It crosses five functions (Identify, Protect, Detect, Respond, Recover) with five asset types: Devices, Applications, Networks, Data, and Users. Plotting capabilities on this grid allows you to see areas where recovery and response receive limited investment. It also clarifies ownership for essential activities such as data integrity checks in AI pipelines.

When the Black Hat MEA team asked Yu what he’s learnt since first developing the Matrix, he said:

“I used to be more rigid in my thinking around how the Cyber Defense Matrix should be interpreted and used. But I realised that I was making it more complex than it needed to be. I would tell my past self that clarity is more valuable than complexity. The most enduring impact comes from making the complex understandable without losing the essence of helping people make better informed decisions based on what the Cyber Defense Matrix shows.” 

Your enterprise (whether you’re working in a critical industry or not) can use this framework to map out your entire security posture and find the elements that need to be strengthened. 

Building resilient systems 

Resilience grows from deliberate preparation. That includes data validation; failover plans that assume networks can (and will) degrade and define how systems continue operating under slower or reduced capacity; and frequent high-level reviews of recovery time objectives, so critical services can be restored as quickly as possible after disruption. 

Gilbert summarised the mindset behind this ongoing preparation work:

“Strength comes from accepting human fallibility and building systems that can withstand it…The blind spot will be a lack of humility; in believing our systems were permanent.”

Resilient organisations build with humility. They expect failure, design accordingly, and keep vital functions alive when events (and machines) move faster than people.