Eighty-five percent of Singapore organisations are already using or piloting AI agents, but all say data challenges have already slowed their AI progress and only 25% are very confident they can detect AI agents operating outside approved parameters.
This is according to a report from Veeam Software, which show that findings from Singapore reveal the sharpest data-trust gap in the APJ region.
As agentic AI moves from pilots into production, Singapore enterprises face an urgent challenge: ensuring that the data powering those systems is visible, governed, secure and resilient.
The research was done between 16 March and 6 April 2026, and involved a global survey of 600 senior executives across industries including financial services, healthcare, manufacturing, retail and technology. Respondents included CEOs, CIOs, CISOs, CDOs and other senior leaders responsible for data, AI, technology and compliance, spanning organizations across North America, Europe and Asia-Pacific.
In Singapore, despite strong executive investment and intent, the ability to control, monitor, and recover from AI failures are critically underdeveloped.
Among executives in Singapore, 70% feel intense or moderate competitive pressure to accelerate AI — with 25% calling the pressure ‘intense’, the highest reading in APJ.
Also, about six in every seven are already using or piloting AI agents but only a quarter are very confident they can detect AI systems operating outside approved parameters (vs. 38% globally).
All Singapore respondents say data challenges have already slowed AI progress (vs. 95% globally).
The figures show a clear trust gap between AI adoption and the governance, visibility, and control required to support it.
‘Shadow AI’
In Singapore, 45% of executives cite ‘your data being used for AI training’ as a top Shadow AI risk — meaningfully higher than the 31% global reading and tied with increased cyber risk (also 45%) as the leading Shadow AI concern locally.
This tracks with how Singapore regulators have approached AI: the Personal Data Protection Act, the Model AI Governance Framework, and the financial regulator’s principles on fairness, ethics, accountability and transparency all put data stewardship at the centre.
As a regional data hub, Singapore’s AI ambitions are inseparable from its data-trust posture. On compliance, 50% of Singapore C-suite respondents flag ‘ensuring personal data is not misused in AI systems’ and a further 50% flag ‘cross-border data transfer restrictions affecting AI systems’ as their single biggest concerns for the next 12 months.
“Most organizations don’t have an AI adoption problem; they have an AI trust problem,” said Anand Eswaran, CEO of Veeam. “With the widespread adoption of autonomous AI agents operating at machine speed, the question transitions from whether you can use AI, to whether you can ensure all your data is secure, governed, compliant and resilient. And should something go wrong, can you recover with precision? That’s how you accelerate safe AI at scale without accelerating reputational and operational risk.”
This combination of rapid AI adoption, coupled with incomplete visibility and unclear accountability creates the conditions for failures that are difficult to detect, explain, and contain.
The research uncovers a significant perception gap between the boardroom and the operational teams responsible for delivering AI outcomes. Progress frequently stalls between intent and execution: governance exists inconsistently, data is managed reactively, and ownership is assigned but fragmented.
Almost half (48%) of CEOs believe trusted, secure, and compliant data could unlock more than 25% revenue growth.
Also, 83% of CEOs feel pressure to accelerate their AI and data capabilities.
This combination of rapid AI adoption, coupled with incomplete visibility and unclear accountability creates the conditions for failures that are difficult to detect, explain, and contain.
When AI fails
As AI systems become more autonomous, the nature of failure is shifting. Risk is moving away from traditional system outages toward data-level failures that are harder to detect, explain, and contain.
The research warns that machine-speed mistakes can outpace detection, forcing resilience to evolve from broad recovery to precision – restoring only what is impacted, rather than rewinding entire environments.
Among organizations running AI today, only a minority could identify within minutes: 29% which systems it accessed; 25% what actions it took; 24% what decisions it influenced and; 22% which data the system used.
Only 40% of leaders are very confident they can isolate and precisely reverse an agentic AI failure.
The governance challenge is converging on data from two directions: internal demand and external scrutiny.
Inside organizations, unauthorized AI use is now mainstream, with 95% reporting unauthorized AI use within their organization and 93% viewing it as a significant risk.
Yet only 25% offer approved alternatives, meaning most are trying to suppress demand rather than govern it effectively, and 44% say increased cyber risk is the top “Shadow AI” risk.
At the same time, regulatory pressure outside the organization is intensifying. 61% of organizations say the EU AI Act has already influenced AI investment strategies in the last 12 months, while 47% cite maintaining audit trails for AI decisions as their biggest compliance concern.
