Singapore organisations report more AI-related security issues than any other market surveyed, according to new global research from Illumio, the breach containment company. The findings show AI emerging as a new kind of insider risk that uses legitimate access to make decisions, reach sensitive data and take actions organisations cannot always see or control.
The research, Licensed to Act: How AI Autonomy Is Outrunning the Limits Placed, was conducted on 1,100 senior IT security decision-makers globally, including Singapore. It finds that 74% of Singapore organisations now delegate decisions and actions that required human approval a year ago, yet only 43% can see and audit every action taken by an AI agent (49% globally). That leaves a clear gap between the authority Singapore organisations give AI and their ability to check what it does.
In Singapore, the risk has already arrived. Nine in 10 organisations (91%) experienced at least one AI-related security issue or warning sign in the past 12 months, the highest of the 10 markets surveyed and well above the 76% global figure. While much of the AI security debate has focused on how attackers might use the technology, the findings also reveal risks emerging from organisations’ own AI systems as they gain access to business applications, data and infrastructure.
Four in 10 (41%) have had an AI agent or tool connect to systems without security approval, the highest of any market and well above the 26% global average. More than a third (35%) say AI produced or triggered an unauthorised action, again the highest of any market and 30% say AI accessed data it should not have reached. Unlike an external attacker, an organisation’s own AI may already hold the credentials and connectivity needed to reach critical systems and data.
AI is also gaining authority inside the security function. Three-quarters (75%) of Singapore organisations use AI within cybersecurity operations (67% globally) and 30% are comfortable letting it run security investigations without human review.
Other key findings include:
- Only 50% are willing to accept more cyber-risk than 12 months ago (56% globally). But among that group, 58% permit autonomous AI actions in production and 44% give AI privileged identities, well above the global figures of 47% and 36%.
- Just 39% are very confident they could contain a compromised or misused AI system before it reached critical systems or data, against 57% globally. And while 72% agree that limiting what AI can access matters more than predicting every threat, only 25% name explicitly defined permitted actions as one of their two main controls.
“Singapore has done more than most countries to set out how Agentic AI should be governed and organisations here are clearly careful about the risk they take on,” said Andrew Kay, Senior Director of Systems Engineering, APJ at Illumio. “But caution alone isn’t keeping AI in its lane. When four in 10 organisations have found AI plugged into systems nobody signed off on, the problem isn’t intent. It’s reach. Trying to predict every move an agent makes is a losing game. Businesses need to decide exactly what each agent can touch, then enforce those limits in the application and across the infrastructure underneath.”
As AI moves from an adviser that recommends actions to an operator with growing authority to take them, the research points to one practical response for Singapore businesses: limit what AI can reach through Zero Trust, rather than trying to predict everything it might do.
