Artificial Intelligence (AI) is changing every aspect of life and work, and revolutionising the world of cybersecurity. But AI is a double-edged sword, and organisations must take control of it to maintain resilience in an AI era.

This is according to Garith Peck, Managing Executive of Cybersecurity at BCX, and Chris Bester, Consultant in Cybersecurity Management & Governance at BCX, speaking on a recent BCX Connects episode.

Accelerated evolution

Peck notes that AI technology has been evolving over 70 years, but its recent evolution has been unprecedented.

“There are 75 years of learning behind AI – from Arthur Samuel’s self-learning checkers programme in 1949, to IBM’s Deep Blue supercomputer beating Garry Kasparov in 1997. Each milestone marked a shift in the programme logic,” he says. “But by the 2010s, deep learning had started taking over and cybersecurity tools had started moving away from static, signature-based detection towards behavioural, anomaly-based models. Today, with the Large Language Models like ChatGPT and DeepSeek, AI can reason, explain and adapt in ways that were once unthinkable.”

“AI has moved from science fiction to boardroom reality. It’s reshaping industries, redefining productivity, and now it’s transforming cybersecurity,” he says.

Both a shield and a weapon

Peck and Bester highlight AI’s tremendous capabilities to transform both cybercrime and cybersecurity.

Peck says: “As AI becomes more capable, it also becomes more unpredictable. It’s both a shield and a weapon, depending on who’s holding it. The dual nature of this technology means that it empowers defenders but also amplifies attacks.”

As a protector, AI automates threat detection and identifies anomalies faster than humans can.

Gartner found that AI-driven analytics can improve threat detection speed by up to 40% and reduce false positives through continuous learning.

Peck notes: “Machine Learning can detect patterns across millions of data points and identify anomalies humans might miss. Predictive analytics can also anticipate vulnerabilities before they’re exploited, improving proactive defences. Automated response is another breakthrough – AI can contain and neutralise threats within seconds. Its scalability is unmatched and it can process data volumes that would overwhelm security teams.”

In the hands of cybercriminals, however, AI is being used to craft more sophisticated attacks.

Says Peck: “On the dark side of AI, it’s estimated that at least 40% of modern cyberattacks use AI in some form, while an MIT study found that 80% of ransomware attacks now use AI. Convincing deepfake social engineering and automated phishing that learns from user behaviour are on the increase. With AI, non-English-speaking threat actors can write perfect phishing mails, so people need to be more on guard.”

He also highlights the risk of bias in AI models trained on unbalanced data, which can distort access and detection outcomes.

“AI security models themselves can be poisoned and manipulated. For example, their datasets can be corrupted, and AI can be taught to misclassify things. It’s the digital equivalent of blinding your guard dog,” he says.

Peck and Bester also emphasise that privacy and data compliance must be maintained as AI scales, with emphasis on traceability and explainability.

Shadow AI risks

Another new risk is the emergence of shadow AI, they warn.

Comments Peck: “We’ve reached a stage where AI is now in every app and workflow, but the challenge is that it’s not all sanctioned, and we are now seeing the rise of ‘shadow AI’. Just like shadow IT before it, shadow AI represents innovation without oversight. Employees experiment with AI tools, build models and automate tasks without formal governance. Shadow AI learns and evolves, and can also expose data and offer unpredictable outputs.”

Bester notes: “It’s a significant challenge for cybersecurity leaders, but we must remember that the battleground hasn’t changed – the weapons have. My worry is that cybersecurity leaders will become complacent and think that AI is going to do it all for them. But AI is a tool. We have to embrace it and recognise that our enemy is also embracing it.”

Defending intelligence

Peck is of the view that cybersecurity needs to shift from defending systems to defending intelligence. “The dual-use nature of AI is perhaps its greatest paradox: the same models that build code securely can also generate exploit kits,” he says. “Leadership must therefore evolve from managing systems to managing intelligence. AI governance must be seen as core to the cyber-strategy and not as a compliance afterthought.”

Responsible use is predicated on strong governance frameworks, and with regulations coming from the EU and others, organisations need to ensure they start being compliant, and stay that way, they advise.

Governance in an AI era

Peck and Bester believe that guardrails, visibility and governance are key to safely adopting AI.

Bester says: “If you ban something like shadow AI, people see it as a challenge to break that ban. The curious minds will always embrace it. Organisations need to put guardrails in place, or even go so far as to create a fenced-off playground for shadow AI. They should create an environment where people can be innovative, but where the organisation has control. To counter external threats, the old adage ‘you have to think like a criminal to catch one’ is still valid. We have to ensure our internal governance is on par so that our AI doesn’t become an insider threat. We must also ensure that the tools are used responsibly, with policies and governance frameworks in place.”

Peck agrees: “Banning AI won’t stop its use. The answer lies in creating boundaries, visibility, and implementing ethical frameworks and controls.”

He adds: “AI is changing everything, including how we govern. It forces us to confront an uncomfortable truth: that innovation is moving faster than regulation. The answer isn’t fear, it’s disciplined regulation. The organisations that win this war will be those that pair the power of AI with strong governance, ethical frameworks and relentless focus on trust.”

Garith Peck and Chris Bester discussed AI in cybersecurity during a Cybersecurity Awareness month podcast on BCX Connects. To listen to the full podcast, go to https://bcx.dev.treemind.solutions/bcx-connects/