AI linked to 83% of breaches, Gigamon survey finds

Gigamon has published a survey finding that AI is now involved in 83 per cent of reported security breaches. It found that 65 per cent of organisations globally and 53 per cent in Australia experienced a breach in the past year.

The results point to a gap between rising spending on cyber security tools and organisations' ability to detect and respond to threats across hybrid cloud environments. Breach rates rose 18 per cent year on year, even as more than nine in 10 organisations continued investing in new security tools and governance measures.

The survey covered more than 1,000 security and IT leaders in Australia, France, Germany, Singapore, the UK and the US. It found that many organisations believe they are managing AI risk effectively even as incident rates remain high, with nearly two-thirds saying their approach to securing new AI technologies is either defined or integrated.

In Australia, the report described a market under pressure from growing AI use, expanding hybrid cloud estates and tighter regulatory scrutiny. It found that 91 per cent of Australian security leaders are recalibrating hybrid cloud risk in response to AI-driven threats, while 86 per cent view metadata as critical to improving visibility across these environments.

The data suggests AI is being used on both sides of cyber security operations. According to the survey, 94 per cent of respondents said AI autonomously initiates security functions without human interaction, most often in alert triage and prioritisation. At the same time, organisations reported a range of AI-related incidents, including external AI attacks, internal leaks, unsanctioned use of AI and direct attacks on large language model systems.

Australian organisations are also seeing breach activity climb. Local breach activity rose 17 per cent year on year, adding to a broader increase over the past three years.

Cloud concerns

The survey also found a shift in attitudes to where AI workloads should sit. Most respondents now see data lakes as more secure for AI workloads, while 70 per cent said they are reluctant to deploy AI in public cloud environments, up from 54 per cent a year earlier.

That trend reflects broader concerns about data visibility and control as businesses increase their reliance on cloud-based systems. In hybrid environments, defenders often struggle to track data in motion across encrypted traffic, East-West traffic and AI workloads, making it harder to spot abnormal behaviour before damage is done.

Among organisations that suffered a breach, only 30 per cent said they had the tools needed to respond effectively. The finding underlines the report's central argument that buying more tools does not necessarily lead to better outcomes if security teams cannot see how systems and data interact across complex estates.

Visibility gap

Gigamon argued that visibility has become the leading security priority in defending against AI-related threats. It said network-derived telemetry, including metadata, packets and flows, can help organisations identify threats earlier by giving analysts a fuller picture of activity across cloud and on-premise systems.

The report found that 93 per cent of respondents believe access to packet-level data and richer application metadata is essential to detecting and understanding modern threats. Board-level interest also appears to be rising, with 90 per cent of leaders saying their boards support investment in deep observability measures.

Shane Buckley, president and CEO of Gigamon, said the research shows how quickly the offensive use of AI is changing the threat landscape. "AI is embedded in nearly every stage of the attack chain, enabling adversaries to outpace detection and response," Buckley said. "While 93 percent of organisations are investing in new security tools, 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."

Another concern highlighted in the survey is the longer-term risk posed by quantum computing. It found that 87 per cent of respondents fear so-called harvest now, decrypt later attacks, in which encrypted data stolen today could be read in the future once quantum techniques become available.

For Australian organisations, those concerns come as regulators place greater emphasis on operational resilience, information security and control over critical systems and data. The survey suggests that, as AI adoption spreads, many security teams are reassessing whether their existing monitoring and response models are keeping pace with the speed and scale of new attacks.

The research was based on an online survey of 1,023 respondents and is the fourth annual edition of Gigamon's hybrid cloud security study.