AI at scale, risk at high speed

Emerging cyber risks shaping business exposure, uncertainty and preparedness

Setting the scene

This year is shaping up to be at least as challenging as last. In fact, nearly half of risk analysts expect 2026 to be more challenging for business than last year, with great power competition as the most likely source of geopolitical risk.[1]

As geopolitical fragmentation continues, a ‘new normal’ of volatility and unpredictability requires deft management by businesses. Increased digital interconnectivity drives complexity, amplifying the speed, scale and impact of potential risk events and scenarios.

Regulation will continue to be fluid and politically driven; businesses will be subject to scrutiny by governments at home and abroad. There are significant differences in regulatory approaches across regions and businesses should be braced for more divergence on technology, data, AI and sustainability, with enforcement increasingly politicised.

Deep interconnectivity between risks also means the impact of specific risk events and scenarios could be felt in a variety of ways and across geographies.

Digital disruption, cyber and AI governance

Digital disruptions remain a top risk to organisations globally in 2026, as growing geopolitical tensions continue to intersect with financially motivated cyber campaigns.[2] New global data reveals that 58% businesses experienced a cyber event in the past 12 months and 57% of those suffered revenue loss as a result.[3]

The wider availability of AI tools will increase the frequency, sophistication and capability of such attacks, enabling faster reconnaissance, easier entry, more convincing social engineering and greater automation of cyber processes. Running parallel, 80% businesses are already using AI in their operations[4] – the mass rollout of AI into core systems, and increasing integration with third party suppliers, further expands available attack surfaces and increases interdependencies.

Recent threat intelligence indicates cyber intrusions are accelerating materially, reducing the time available to detect and contain incidents from hours to minutes. Attackers are using widely available AI tools to scale phishing, reconnaissance, and evasion, while also targeting AI and cloud environments directly. The net effect is higher likelihood of rapid business disruption, data loss, and financial impact unless identity, cloud, and exposure management controls keep pace. As a result, 72% businesses expect cybersecurity budgets to increase and with 37% specifying a rate beyond inflation.[5]

Case Study: AI-accelerated cyber intrusions

2025 trends reported by CrowdStrike across 280+ tracked adversaries:
• Intrusions are faster: average breakout time reported at 29 minutes (fastest observed 27 seconds)
• Data can leave quickly: within four minutes in one case
• Growth in AI-enabled activity: 89% increase year-on-year, with AI used for reconnaissance, credential theft, and evasion
• Cloud and identity pathways are central: cloud-conscious intrusions up 37% and cloud targeting/intelligence collection up 266%
• Social engineering volume rises: 563% increase in incidents using fake CAPTCHA lures and a 141% increase in spam emails [6]
• Financial impact is high: a single reported crypto theft was valued at USD $1.46B

Source: CrowdStrike: Global Threat Report [7]

Meanwhile, governments are playing catch up with both threat actors and technology developments, leading to a flurry of new regulation across cyber security, data protection and AI governance.[8] Regulations differ greatly across global regions with notable developments in ransom payment oversight and reporting, AI usage in critical industries and ongoing expansion of data categories that may be subject to regulatory penalties. This creates a complex and fast evolving compliance environment, especially for businesses with multinational operations.

The growing use of agentic AI [10] is creating new opportunities for organisations to improve efficiency and automate complex workflows. However, it also introduces new forms of exposure, making unprotected businesses an attractive target – and tool – for cyber threat actors.

The shift to agentic AI marks a structural change in AI adoption and deployment for businesses. Unlike the previous, more passive tools, agentic systems can act autonomously, making complex decisions in operational processes. As these systems become more deeply embedded in day-to-day operations and workflows, the potential impact of disruption increases.

[11] [12] [13]

Threat actors are increasingly using AI agents to speed up and scale cyber attacks, allowing them to operate beyond the capabilities of humans. In the short term, this is likely to benefit threat actors who already have access, infrastructure and intent to carry out complex attacks, including state-linked groups and highly organised criminal networks.

However, as AI tools become easier to access and use, more actors across the threat landscape are likely to adopt them, lowering barriers to entry and increasing both the speed and volume of attacks.

CASE STUDY: Nation-state use of agentic AI highlights emerging AI-driven espionage capabilities

In November 2025, Anthropic reported on an espionage campaign attributed to a threat actor likely operating from China, which made extensive use of the Claude Code tool. [14] The activity targeted around 30 organisations worldwide, in sectors including technology, financial services, manufacturing and government.

According to Anthropic, the actor used Claude’s agentic AI capabilities to autonomously carry out 80–90% of tactical actions against designated targets, with human input largely confined to higher level decision making. This allowed the attack to be conducted faster than would typically be possible using human operators alone.

However, Anthropic also found that the AI system, Claude, generated inaccurate information and regularly overstated its findings, which are likely to have dampened the effectiveness of the incident and slowed progress against some targets.

Lower capability and/or opportunistic threat actors will continue to experiment with widely available AI tools to target organisations through financial fraud and misinformation campaigns. Sectors such as financial, IT and telecommunications, and government remain among the most frequently targeted for such attacks (see Figure 2). AI generated deepfakes are being used to spread false narratives that undermine trust in organisations and damage reputations.

Meanwhile, these tools are enabling more direct financial crime. For example, deepfake video or audio messages impersonating senior executives can be used in payment diversion campaigns, in which employees are tricked into transferring funds to accounts controlled by attackers.