The AI cyber divide

How organisations set themselves up for success when both defenders and attackers are accelerating.

The AI cyber divide

Artificial intelligence is compressing time and time has become the new battlefield.

AI is helping organisations innovate faster, automate decisions, improve productivity, and unlock new growth opportunities. At the same time, AI is helping threat actors identify vulnerabilities, scale social engineering, automate reconnaissance activities, generate malicious code, and adapt attacks at unprecedented speed.

The result is a growing divide. On one side organisations are using AI to create competitive advantage. On the other side organisations are struggling to govern it, secure it, and understand where it’s creating new exposure.

The challenge isn’t that AI is creating new risks. The challenge is that AI is magnifying existing weaknesses much faster than most operating models can adapt and remediate.

What once took months now takes days. What once took days now takes hours. What once required specialist expertise is increasingly accessible to anyone with the right tools.

Success in the future of cybersecurity will belong to organisations who can move at AI speed without losing control.

The question isn’t whether AI will transform your business. 

The question is whether your cyber strategy will evolve quickly enough to keep pace with the transformation.

Security can no longer be bolted on

Many organisations still govern cyber as a technology function and AI as an innovation initiative. That separation is becoming dangerous. AI decisions influence customer outcomes, operational processes, employee productivity, financial reporting, supply chains, and strategic decision-making. Cybersecurity has to become a foundational component of AI adoption, not a checkpoint at the end of a project.

Future-ready organisations build security, privacy, resilience, governance, and trust into AI initiatives from the beginning.

  • What data can AI access?
  • What decisions can AI influence?
  • Who is accountable?
  • How is AI monitored?
  • What happens when AI fails?

If those questions can’t be answered quickly, the organisation is moving faster than it can govern.

Identity becomes the new control plane

The future attack surface looks dramatically different from today's. Human identities are no longer the only concern. Service accounts, machine identities, application programming interfaces (APIs), automation platforms, AI assistants, autonomous agents, and third-party integrations are expanding rapidly. Many already possess access privileges equivalent to employees.

Organisations who succeed will treat identity as the control plane of the digital enterprise.

  • Who has access?
  • What has access?
  • Why does access exist?
  • Is access still required?
  • How is behaviour changing over time?

Visibility into identities, permissions, and privileged access will become more valuable than visibility into networks alone. 

You can’t protect what you can’t identify and what you can’t identify may not even be human.

Shift from managing incidents to managing exposure

Most cybersecurity programmes were designed for a time where change was slower. That world no longer exists. Waiting for annual risk assessments, quarterly reviews, or periodic testing creates blind spots that AI-powered threats can exploit.

Leading organisations are moving toward continuous exposure management.

  • Replace “Are we compliant?” with “Where can we be exploited today?”
  • Measure outcomes rather than activity
  • Continuously evaluate critical assets, sensitive data, identities, vulnerabilities, third parties, AI systems, and business dependencies.

The goal isn’t perfect security. The goal is reducing uncertainty before it becomes disruption.

Use AI aggressively. Govern it relentlessly.

Some organisations will hesitate because of risk. Others will accelerate without adequate controls. Neither approach is sustainable. The organisations who lead will embrace disciplined ambition. They’ll use AI extensively to improve security operations, automate investigations, enrich alerts, identify emerging threats, reduce manual effort, and improve decision-making.

But they’ll also establish clear guardrails.

  • Retain human accountability for high-impact decisions
  • Define thresholds for autonomous actions
  • Monitor, test, challenge, and validate AI-generated outcomes
  • Require evidence that AI is secure, reliable, explainable, and aligned to business objectives.

Trust is built through evidence, not optimism. The winners won’t be organisations who deploy the most AI. They will be organisations who can prove their AI deserves to be trusted.

Build resilience for machine-speed failure

Many resilience plans were built for a slower time, and when mistakes occurred at human-speed. The future requires planning for machine-speed failures. A compromised AI model, poisoned dataset, manipulated algorithm, rogue autonomous process, or exploited AI agent can create consequences at a scale and pace traditional response models will struggle to contain.

This means resilience must evolve.

  • Test AI failure scenarios
  • Incorporate AI into crisis exercises
  • Validate recovery procedures
  • Prepare for compromised models and corrupted data
  • Establish clear escalation paths and accountabilities. 

Resilience isn’t just about recovering systems. It’s about maintaining trust when automated decisions, capabilities, and digital ecosystems fail.

The leadership challenge

Technology will matter. Controls will matter. Platforms will matter. But leadership will matter more. The future of cybersecurity is becoming less about managing technology and more about managing speed, accountability, and trust. Boards and executives have to be able to see where AI is creating value, where it’s introducing exposure, who owns the risk, and whether controls are working.

The organisations who thrive will make cyber and AI governance part of strategic decision-making rather than operational reporting. They will treat cybersecurity as an enabler of growth, not simply a cost of protection. Most importantly, they will recognise that AI isn’t changing the mission. It is changing the pace.

The path forward

The organisations who succeed won’t be those with the most controls, the largest security teams, or the longest risk registers. They’ll be the organisations who can innovate quickly, govern intelligently, adapt continuously, and prove they can be trusted.

The practical priority is to turn that ambition into an operating rhythm. Start with these seven steps that create visibility, accountability, speed, and evidence.


PRACTICAL MOVEACTION AND OUTCOME
1. Map the AI-enabled enterprise

Create a current inventory of AI use cases, embedded AI features, models, agents, data sources, integrations, vendors, and machine identities. Classify each by business impact and exposure.

Outcome: A single view of where AI is operating, what it can touch, and which uses matter most.

2. Name accountable ownersAssign a business owner, technology owner, data owner, and risk owner for every material AI use case. Define who approves deployment, accepts risk, monitors performance, and can stop the capability.

Outcome: No material AI operates without clear decision rights and escalation paths.

3. Create responsible speed

Create a fast lane for low risk uses and stronger gates for AI that affects customers, employees, critical operations, sensitive data, or consequential decisions.

Outcome: Governance accelerates safe innovation instead of treating every use case the same.

4. Secure identities, data and agents

Apply least privilege, strong authentication, secrets management, data access controls, logging, and lifecycle management to both human and non-human identities.

Outcome: Every person, service, model, and agent has only the access it needs, for only as long as it needs it.

5. Modernise defence with AI

Use AI to enrich alerts, improve correlation, accelerate investigation, prioritise exposure, and automate repeatable, high-confidence response actions. Keep accountable human judgment where actions could create material business impact.

Outcome: Detection, investigation, and containment move closer to the speed of the threat.

6. Test continuously

Continuously validate controls, attack paths, model behaviour, recovery procedures, and third-party dependencies. Add AI compromise, data poisoning, agent misuse, and automated decision failure to tabletop exercises.

Outcome: Leaders receive evidence that controls and resilience work under realistic pressure.

7. Measure what matters

Replace activity-heavy reporting with a concise view of material exposure, time to detect and contain, privileged access, control effectiveness, resilience, AI exceptions, and risk acceptance.

Outcome: Boards and executives can see whether risk is reducing and whether the organisation can move with confidence.

Avoid attempting to perfect all seven steps before beginning. Start with the material AI uses, critical business services, sensitive data, privileged identities, and most credible attack paths. 

Establish a 90-day action plan, identify the owners, agree upon the evidence and reports leaders will see, and review progress at a fixed cadence.

AI is creating one of the largest shifts cybersecurity has ever faced. It is also creating one of its greatest opportunities.

The future belongs to organisations who move at machine speed without sacrificing human judgment, accountability, and trust.

AI will reward the prepared and expose the complacent. 

The real competitive advantage won’t be adopting AI first. 

It’ll be governing it better than everyone else.

CONTACT US

Key Findings:

  • AI accelerates both business growth and cyberattacks. Existing vulnerabilities can become serious threats faster than organisations can respond.
  • Cybersecurity must be built into AI adoption from the start. Data protection, access controls and risk management should be considered at the planning stage.
  • Access management is becoming critical. Organisations need visibility into the permissions of employees, systems, AI agents and automated services.
  • Incident response must evolve into continuous exposure management. Ongoing monitoring, control testing and attack scenario assessments help identify weaknesses before they cause disruption.
  • Responsible AI use creates a competitive advantage. Clear accountability, human oversight and evidence-based validation enable innovation without compromising control or trust.

Subscribe to BDO in Ukraine Newsletters.

SUBSCRIBE

Key Contact

Andrii Borenkov

Andrii Borenkov, CFA

Partner, Head of Advisory
View bio