Artificial intelligence (AI) has been involved in 84 percent of reported security breaches in Singapore over the past 12 months, enabling attackers to operate with greater speed and scale than many organizations can defend against, a survey showed Wednesday.
Gigamon, which released its 2026 Hybrid Cloud Security Survey, revealing a fundamental shift in the cyber threat landscape.
Despite expanded investments in tools and governance policies, 65 percent of organizations globally have experienced a breach in the past year, reflecting an increase of 40 percent over the past three years.
The findings highlight a growing imbalance as adversaries leverage AI to accelerate cyber attacks, while defenders are constrained by fragmented visibility into what’s happening across their networks.
“AI is embedded in nearly every stage of the attack chain, enabling adversaries to outpace detection and response,” said Shane Buckley, President and Chief Executive Officer at Gigamon.
While 91 percent of Singaporean organizations are investing in new security tools, he noted many still lack visibility into how data moves across their environments, creating confidence without control.
“Closing this gap requires deep observability, giving security teams the clarity needed to detect threats earlier and respond with precision,” he added.
The annual study, now in its fourth year, surveyed more than 1,000 global Security and IT leaders across Australia, France, Germany, Singapore, the United Kingdom, and the United States.
The survey report, “Reality Check: Exposing the AI Security Illusion,” explores the disconnect between confidence and reality, revealing a critical blind spot in how organizations assess AI-driven risk.
According to the survey, confidence is outpacing capability. Nearly two-thirds (63 percent) of Singaporean organizations believe their ability to secure new AI technologies is “defined” or “integrated,” yet 60 percent experienced a breach in the past year, and 1 in 3 experienced multiple breaches.
Meanwhile, AI is transforming both sides of the equation. AI is now embedded across security operations in Singapore, with 93 percent reporting it autonomously initiates security functions without human interaction, most commonly in alert triage and
prioritization (51 percent).
At the same time, AI security incidents span multiple risk categories, including: external AI attacks (41 percent globally); internal leaks (30 percent globally); unsanctioned use of AI (30 percent globally); direct attacks on LLM systems (33 percent globally).
The survey also showed trust in cloud AI deployments is continuing to erode.
As risk increases, data strategies are shifting. Most leaders globally (72 percent) now believe data lakes are more secure for AI workloads, compared with 70 percent that say they’re reluctant to deploy AI in public cloud environments, up from 54 percent the previous year.
Quantum computing risk is also accelerating the timeline. Looking ahead, 92 percent of Singaporean leaders, the highest across the regions, fear “harvest now, decrypt later” attacks, putting today’s encrypted data at future risk and underscoring the longer-term implications of current visibility gaps
Meanwhile, visibility emerged as the top security priority in defending against AI threats in this year’s study, yet it is also where defenders are falling behind.
As attackers leverage AI to move faster and operate at scale, organizations still lack a complete view of data in motion across encrypted and East-West traffic, AI workloads, and cloud environments.
Among those that experienced a breach, only 30 percent say they had the tools needed to respond effectively, highlighting a critical gap between investment and outcome.
To close this gap, organizations are turning to deep observability.
By using network-derived telemetry, including metadata, packets, and flows, and feeding it into security, observability, and cloud tools, organizations can gain complete visibility into all data in motion.
Nearly all (97 percent) of Singaporean leaders agreed, reporting that access to packet-level data and rich application metadata is essential to detecting and understanding modern threats, going beyond the visibility that traditional MELT data provides today, said the survey.
This shift is also reaching the boardroom, with 94 percent of Singaporean leaders reporting that their boards now support deep observability initiatives, signaling an ongoing commitment to modern approaches that can defend against today’s AI-driven
threats.