Only 53% of Malaysian Organisations Confident They Can Respond to Compromised AI Agents

KUALA LUMPUR: Malaysian organisations are struggling to keep pace with the speed of AI-driven cyber threats, with only 53% saying they are highly confident they could detect and respond to a compromised AI agent, according to new research from Cisco.

The finding comes as 83% of organisations in Malaysia reported experiencing a material, business-disrupting cyber incident over the past 12 months, with almost one in three of those incidents involving an AI-enhanced attack.

Cisco's 2026 Relentless Defense Report suggests that the challenge is increasingly organisational rather than purely technological.

The research found that organisations need to combine broad security coverage with faster detection and response, while reducing the internal friction that can prevent security teams from acting quickly. Malaysia recorded an average score of 65 out of 100 in the report's assessment of cyber defence capabilities.

Only 8% of organisations in Malaysia qualified as "Relentless Defenders", the report's highest-performing group, a proportion that was also 8% globally. Cisco Malaysia Managing Director Chee Kheong Lee said the growing use of AI had changed the speed at which organisations needed to respond to cyber threats.

“AI has fundamentally raised the stakes for cyber defence,” he said.

He said organisations that wanted to stay ahead would need to embed security into their operations, rather than treating it primarily as a compliance exercise, while reducing silos that prevent teams from making decisions quickly.

Organisational friction slows response

The report found significant gaps in coordination among Malaysian security, networking and IT teams. Only 43% of organisations said these teams operate as a single team with shared visibility and tools. The problem also extends to how cybersecurity information is communicated to senior management.

Around 29% of respondents said the metrics they measure are too technical to persuade executives, while 43% said they struggle to translate business metrics into information that can influence decision-making. A further 34% said they had no consistent way of measuring security performance.

Data fragmentation is another obstacle. Almost half, or 49%, of Malaysian respondents said they spend more time collecting and correlating information across systems than dealing with the actual threat. Another 40% said critical insights can be missed because relevant data remains inaccessible.

The findings point to a growing disconnect between the speed of emerging threats and the speed at which organisations can make decisions internally.

AI speeds up security but raises new risks

AI is nevertheless delivering measurable benefits for security teams. More than eight in 10 security practitioners surveyed reported improvements of at least 25% in threat detection and response, while more than one in three said AI was significantly reducing their operational burden. However, extensive adoption of AI-powered security capabilities remains below half of organisations.

Deployment is most advanced in areas including network security, data security and privacy, and security operations, while identity and access management has seen lower levels of extensive adoption.

More than one-third of respondents said they had extensively deployed AI agents in security operations. But concerns around data exposure, compromised systems and third-party handling are slowing wider adoption. The result is a paradox for organisations: AI can accelerate cyber defence, but the same technology introduces new risks that security teams must be capable of identifying and controlling.

The defenders that move faster

Cisco's highest-performing organisations stand out less for simply spending more on technology and more for how they organise their security operations.

Among the Relentless Defenders globally, 83% said they operate as a single coordinated security team with shared visibility and tools across security, network and infrastructure, compared with 37% of other defenders.

They were also substantially more likely to apply consistent controls to AI agents and non-human identities, at 86% compared with 37%.

Confidence in responding to compromised AI agents was correspondingly higher, with 77% of Relentless Defenders saying they were very confident they could detect and respond to such an incident, compared with 51% among other defenders.

The report also found that 25% of Relentless Defenders said AI had improved threat-detection speed by more than 100%, compared with 7% of other organisations. The difference was also visible in incident outcomes.

Among organisations that increased their security spending, 71% of Relentless Defenders reported a decline in incidents, compared with 39% among other organisations. The findings suggest that cybersecurity performance is increasingly dependent on how effectively technology, people, data and decision-making processes operate together.

The Cisco report is based on a double-blind survey of 8,000 security professionals across 30 markets, conducted in April and May 2026. Independent research firm Satori Experience conducted the analysis, assessing organisations across three areas: coverage, speed and friction.

For Malaysian businesses increasingly deploying AI across their operations, the findings point to a broader challenge: having the technology to detect a threat may no longer be enough if the organisation cannot decide and act quickly when the threat emerges.