AI & Emerging Threats

A new generation of cyber threat

Artificial intelligence is transforming the way organisations operate, but it is also changing the cyber-security landscape.

Many of the techniques used in cyber attacks are not new. Phishing, social engineering, credential theft, malware and vulnerability exploitation have existed for years. What AI changes is the speed, scale and sophistication with which these techniques can be used.

Activities that previously required significant time, expertise or resources can increasingly be automated. Attackers can analyse information, identify potential targets, create convincing content and adapt their approach more quickly than before.

The result is not simply a new category of cyber attack. It is the potential industrialisation of existing cyber threats.

Social engineering becomes more convincing

People have always been an important target for cyber attackers.

Generative AI makes it significantly easier to produce convincing emails, messages, documents and other communications. Poor spelling and obviously suspicious wording can no longer be relied upon as indicators of a phishing attempt.

AI can also help attackers analyse publicly available information about an organisation and its employees, allowing attacks to become more personalised.

Voice cloning and synthetic video add another dimension. An attacker may attempt to impersonate a colleague, executive, supplier or other trusted individual through increasingly convincing digital communications.

This means organisations need to reconsider security processes that depend heavily upon human recognition or trust.

Attacks at machine speed

Automation has long been part of cyber crime, but AI has the potential to make automated attacks considerably more adaptive.

AI-enabled tools can assist with analysing systems, identifying potential weaknesses, processing large quantities of information and adapting activity according to the results obtained.

This creates an important change in the economics of cyber attacks.

An attack that would once have required significant human effort may potentially be conducted against many more organisations simultaneously. Smaller organisations and lower-value targets can therefore become worthwhile targets simply because the marginal cost of attacking them is reduced.

Cyber resilience must consequently consider not only increasingly sophisticated attacks, but potentially far greater volumes of attacks.

AI creates risks inside the organisation

The threat does not come solely from attackers.

Organisations are rapidly adopting generative AI tools, AI assistants and increasingly autonomous systems. Each can create new security considerations.

Employees may inadvertently provide sensitive corporate information to external AI services. AI applications may be granted access to documents, email, databases and business systems. Models and their supporting infrastructure may themselves become targets. AI-generated information can also be inaccurate or manipulated.

As AI becomes integrated into business processes, organisations therefore need to understand what information AI can access, what actions it can perform and who remains accountable for the outcome.

Traditional access-control principles remain important, but they must increasingly account for machines acting on behalf of people.

AI introduces new attack surfaces

AI systems can themselves become part of the attack surface.

For example, systems using large language models may need to process information from users, documents, websites and other external sources. Malicious instructions hidden within that information can potentially influence how an AI system behaves.

The risks become more significant as AI moves from simply providing information to being able to take actions.

An AI assistant that can only answer a question presents one level of risk. An AI agent able to access corporate systems, retrieve sensitive information, modify data or initiate business processes presents a very different one.

Organisations therefore need security architectures that consider the identity, permissions, data access and autonomy of AI systems alongside those of traditional users and applications.

Telecommunications and critical infrastructure

These developments are particularly significant for telecommunications and other critical infrastructure.

Telecommunications networks combine large-scale infrastructure, software, cloud platforms, operational systems, suppliers and extensive interconnected technology. They also provide services upon which other organisations and critical services depend.

AI can provide powerful capabilities for managing and defending these environments, including identifying anomalies and analysing large volumes of security information.

But attackers have access to many of the same technological advances.

For organisations responsible for important infrastructure, the challenge is therefore twofold: using AI safely while preparing to defend against adversaries that increasingly use AI themselves.

People, process and technology must evolve together

AI-enabled cyber threats cannot be addressed by purchasing another security product.

People need to understand how AI changes social engineering, information handling and decision-making.

Processes need to define how AI is approved, governed, monitored and used, including appropriate controls when AI systems are permitted to perform actions.

Technology needs to provide appropriate identity, access, monitoring, data protection and security controls around both conventional systems and emerging AI capabilities.

These elements need to evolve together.

Introducing powerful AI technology without the corresponding organisational controls can create new vulnerabilities. Equally, attempting to control AI solely through policy without appropriate technical enforcement is unlikely to provide lasting protection.

Preparing for what comes next

The development of AI is moving quickly, and its long-term impact on cyber security cannot be predicted with certainty.

Organisations therefore need an approach based on adaptability and resilience rather than attempting to predict every future attack.

That means understanding emerging threats, continually reassessing risk, controlling how AI is introduced into the organisation and ensuring that security transformation keeps pace with technological change.

Cyber Me helps organisations examine these challenges through the combined perspective of cyber security, telecommunications and organisational transformation, aligning people, process and technology for an increasingly AI-enabled environment.

Protecting today. Preparing for tomorrow.

The question is no longer whether AI will influence cyber security. It already does.

The challenge is ensuring that organisations can take advantage of its opportunities without allowing the speed of technological change to outpace their ability to secure it.