AI’s benefits are contingent on disciplined execution. Poorly implemented solutions can introduce new risks – misconfiguration, biased decision-making, over -reliance on automation and susceptibility to adversarial manipulation – unless organizations embed robust guardrails, security-by-design practices and continuous monitoring. The implication is clear: AI can improve cybersecurity, but only when deployed within sound governance frameworks that keep human judgement at the centre. At the same time, too many controls can create friction, so it is essential to strike a careful balance. Security of AI: from awareness to action According to the Global Cybersecurity Outlook 2026 survey, 87% of respondents identified AI-related vulnerabilities as the fastest-growing cyber risk over the course of 2025. Perception of increase or decrease in cyber risks over the past year FIGURE 8 Increase (%) Neutral (%) Decrease (%)In the past year, do you think the following cyber risks have increased, decreased or stayed the same? 20% 40% 60% 80%100%Denial-of-service attacks28% 18% 54%Insider threat32% 61%Ransomware attack54% 39%Exploitation of software vulnerabilities58% 39%Supply chain disruption65% 32%Cyber-enabled fraud and phishing77% 21%AI vulnerabilities87% 13% 0% Data leaks associated with genAI (34%) and the advancement of adversarial capabilities (29%) stand out as leading concerns for 2026. This marks a striking reversal from previous years – in 2025, advancement of adversarial capabilities topped the list at 47% compared to only 22% for data leaks associated with genAI. The shift underscores a turning point in the AI risk landscape for the upcoming year: while the “AI arms race” between attackers and defenders continues to intensify, attention is pivoting from purely offensive innovation with AI towards the unintended exposure and misuse of sensitive data through generative and agentic systems. Global Cybersecurity Outlook 2026 19