As organisations race to implement AI to keep pace with increasingly sophisticated threats, most remain unwilling to trust AI with autonomous security decisions, leaving security leaders caught between the need for machine-speed defenses and the demand for trusted outcomes, according to a report from Arctic Wolf.
This is based on a survey of 1,350 security and IT decision-makers across the world, including in Singapore, conducted by Sapio Research.
The report found that AI has quickly become a prerequisite for modern security operations as 94% of organisations now use LLMs, while also 94% say AI capabilities influence cybersecurity purchasing decisions, and more than half (51%) consider AI functionality a requirement when evaluating vendors.
Despite that enthusiasm, only 14% have made AI central to their security operations strategy, underscoring a lack of confidence in autonomous AI.
Just 53% trust AI to perform even narrowly defined security actions, such as blocking malicious IP addresses at a firewall for fear of these actions being triggered from a false positive detection.
In Singapore, the findings reveal a particularly interesting tension. Organisations here are among the world’s most willing to trust AI to support security operations, with more than half trusting AI to triage alerts and patch vulnerabilities independently.
However, this trust has not translated into stronger cyber resilience. Three in every five organisations experienced a major incident this year, with some facing disruption lasting the better part of a year.
The trust gap comes at a time when organisations are facing persistent cyber risk. Nearly two-thirds (63%) of respondents reported experiencing a significant cybersecurity incident within the past year, and almost half (48%) of affected organisations suffered disruptions lasting two weeks or longer.
Despite those outcomes, 96% of leaders remain confident their teams can keep pace with the volume and complexity of today’s threats, highlighting a growing gap between perceived readiness and real-world results.
“What stands out about Singapore is that organisations here trust AI more than almost anywhere else in the world, and they’re also more clear-eyed about its risks,” said Steve Hunter, Arctic Wolf director of engineering in Asia-Pacific.
“That combination of confidence and caution is the right instinct, but it hasn’t yet translated into stronger outcomes,” said Hunter. “The priority now isn’t adopting agentic AI faster, it’s building the oversight, governance and human expertise around it so that trust is earned rather than assumed.”
Findings also show that AI dominates both the risk list and the priority list. AI, LLMs and agentic AI — and their privacy implications — rank as Singapore’s single leading cybersecurity concern, and 53% cite secure AI adoption and data transformation as their top cybersecurity priority, compared with 41% globally.
Also, trust in autonomous AI runs far ahead of the rest of the world. 60% of Singapore organisations trust agentic AI to independently triage alerts and generate reports, and 56% trust it to patch vulnerabilities without human oversight; both well above the 42% global average.
In addition, high trust hasn’t translated into better protection. 61% of Singapore organisations experienced a major cyber incident, roughly in line with the global figure, but 13% report severe disruption lasting six to nine months. This is more than three times the global rate.
Further, the skills shortage is pushing organisations toward outside help. 52% of Singapore organisations are turning to managed services to close the gap, significantly above the 39% global figure.
“Organisations have already decided that AI will play a central role in cybersecurity. The challenge now is trust,” said Adam Marrè, CISO at Arctic Wolf.
“Security leaders want the speed, scale, and efficiency AI can deliver, but they also need confidence that the outcomes are trustworthy, explainable, and aligned to their business,” said Marrè. The future of security operations will belong to organisations that combine trusted AI with human expertise to move faster, make better decisions, and build greater resilience.”
