Cisco warns Singapore firms struggle to spot compromised AI

Only 38% of organisations in Singapore are confident they can detect and respond to a compromised AI agent, according to Cisco. Its latest survey also found that 89% suffered a material, business-disrupting cyber incident over the past year.

The Singapore results reflect a wider regional picture: across Asia Pacific, Japan and China, only 52% of organisations said they were confident they could detect and respond to a compromised AI agent.

The figures come from Cisco's Relentless Defence Report, a survey of 8,000 security professionals across 30 markets. The study assessed organisations on the breadth of their security coverage, the speed of response and remediation, and the operational friction that slows teams down.

In Singapore, the average score was 63 out of 100. Just 6% of organisations qualified for the top-performing group Cisco calls Relentless Defenders, compared with 8% globally.

The report suggests many organisations are being held back less by a lack of security tools than by internal bottlenecks. In Singapore, only 39% of respondents said security, networking and IT teams operated as one team with shared visibility and tooling.

That fragmentation also affects reporting and decision-making. Cisco found that 42% of organisations in Singapore said their metrics were too technical to persuade executives, while 34% struggled to translate business metrics into information that could influence decisions. Another 30% said they had no consistent way to measure security performance.

Data access remains another constraint. Some 37% of respondents in Singapore said they spent more time collecting and correlating data across systems than addressing threats, while 41% said critical insights were missed because data was out of reach.

AI pressure

The report found that AI is increasing the speed and scale of cyber threats even as organisations try to use it in their own defences. In Singapore, more than two in five business-disrupting cyber incidents involved an AI-enhanced attack.

At the same time, many security teams said AI had delivered measurable gains. Nearly eight in 10 practitioners in Singapore reported detection and response improvements of at least 25%, and one in five said AI was significantly reducing operational burden.

Even so, broad adoption remains limited. Fewer than half of organisations in Singapore have extensively deployed AI-powered security tools. Adoption was highest in security operations detection and network security, and lowest in AI workloads, models and systems.

Just over a quarter of organisations in Singapore said they had extensively deployed AI agents in security operations. Respondents cited concerns including data exposure, and limited visibility into and access to AI systems.

The regional data points to similar issues. Across Asia Pacific, Japan and China, more than one in three business-disrupting cyber incidents involved an AI-enhanced attack, while 91% of organisations said they had suffered a material cyber incident in the past 12 months.

Across the region, only 39% of organisations said security, networking and IT teams operated as one team with shared visibility and tooling. Another 42% spent more time collecting and correlating data than addressing threats, and 41% said key insights were missed because data was inaccessible.

Top tier

The report identifies a small group of organisations that appear to be outperforming their peers. Among those that increased security spending, the top-performing group was nearly twice as likely to report falling incident levels, with 71% saying incidents had declined, compared with 38% in Singapore and across the wider region.

Globally, 83% of those top performers said they operated as a single coordinated security team with shared visibility and tools across security, network and infrastructure, compared with 37% of other organisations. The report also found that 86% of this group applied consistent controls to AI agents and non-human identities, compared with 37% of other defenders.

Confidence levels were also markedly different. Some 77% of the top-performing organisations said they were very confident they could detect and respond to a compromised AI agent, compared with 51% of others worldwide.

Tay Bee Kheng, President, ASEAN, Cisco, said the challenge was no longer confined to technology choices. "AI has fundamentally raised the stakes for cyber defense. In a post-Mythos era, the speed of defense is driven by operational agility, not technology alone. The organizations that stay ahead will be those that embed security into how they operate, treat it as more than a compliance exercise, and break down operational silos so the right people can make the right decisions at the right time," Kheng said.

Ben Dawson, President, Asia Pacific, Japan and Greater China, Cisco, said the wider region faced a similar challenge. "Boards have moved past asking whether AI will deliver value. The urgent question leaders are now facing is whether their digital foundation is secure and resilient enough to deploy it. Organisations that treat security merely as a compliance exercise will struggle, while those embedding defense directly into how they operate are the ones setting the pace for innovation in the AI era," Dawson said.