Singapore organisations report more AI security issues than any other market surveyed, according to new research from Illumio, the breach containment company. The study, titled Licensed to Act, points to AI systems that already hold legitimate access acting in ways security teams cannot always see or control.
Key Takeaways
- 91% of Singapore organisations had at least one AI-related security issue or warning sign in the past 12 months, against 76% globally.
- 41% found an AI agent or tool connected to systems without security approval, versus 26% globally.
- Only 39% are very confident they could contain a compromised AI system before it reached critical systems, against 57% globally.
AI security issues hit 91% of Singapore firms
The study was conducted by Sapio Research on behalf of Illumio. It surveyed 1,100 senior IT security decision makers across 10 markets, including 100 in Singapore. Global figures are unweighted averages.
Nine in 10 Singapore organisations (91%) had at least one AI-related security issue or warning sign in the past 12 months, the highest of the 10 markets and well above the 76% global figure.
Authority is moving to AI quickly. Some 74% of Singapore organisations now delegate decisions and actions that needed human approval a year ago, but only 43% can see and audit every action an AI agent takes. The global figure for that visibility is 49%.
Unapproved AI connections lead the findings
Some 41% of Singapore organisations 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. Separately, 35% said AI produced or triggered an unauthorised action, also the highest, while 30% said AI accessed data it should not have reached.
Illumio describes AI as a new kind of insider risk. Unlike an external attacker, an organisation’s own AI may already hold the credentials and connectivity needed to reach critical systems and data.
AI takes on more security work
Three-quarters (75%) of Singapore organisations use AI within cybersecurity operations, against 67% globally. About 30% are comfortable letting it run security investigations without human review.
Caution on risk does not carry over to access. Only 50% of Singapore respondents are willing to accept more cyber risk than 12 months ago, compared with 56% globally. Among those that are, 58% permit autonomous AI actions in production and 44% give AI privileged identities, against 47% and 36% globally.
Zero Trust limits proposed for AI agents
Only 39% of Singapore organisations are very confident they could contain a compromised or misused AI system before it reached critical systems or data, compared with 57% globally. 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.
“When four in 10 organisations have found AI plugged into systems nobody signed off on, the problem isn’t intent. It’s reach,” said Andrew Kay, Senior Director of Systems Engineering, APJ at Illumio.
Kay said businesses should decide exactly what each agent can touch, then enforce those limits in the application and across the infrastructure underneath.
The Infocomm Media Development Authority‘s Model AI Governance Framework for Agentic AI already asks organisations to limit the tools and systems agents can reach. Illumio’s research points to Zero Trust as the practical response, rather than trying to predict everything an agent might do.
