New research from Sumsub and the Singapore FinTech Association finds a widening gap between how much organisations rely on AI and their ability to evidence what it does.
Across Asia-Pacific, AI has moved past assisting people into acting for them, opening tickets, moving money, and completing multi-step tasks with limited human input.
A new benchmark from Sumsub, produced with the Singapore FinTech Association and based on a survey of 720 senior professionals across nine markets, measures how ready organizations are to govern that shift.
The clear verdict found from the survey is that confidence is high amongst all nine markets, however proof of readiness is still scarce.
The report scores organisations from 0 to 100 across three pillars. These pillars are autonomy, responsibility, and traceability. Autonomy and responsibility both sit close to 70. Traceability drops to 61 and is the weakest pillar in every market and sector. Sumsub calls the gap the Accountability Asymmetry.
The confidence organisations have in their ability to explain AI decisions is not matched by their ability to show how those decisions were made. While 95% of respondents say they are confident they can explain an AI decision, only half can reconstruct the decision pathway, and just 38% keep a tamper-proof audit trail.
That gap becomes more important as AI is given greater freedom to act. 60% of organisations are already running AI that can carry out multiple steps with limited human input, while 63% have experienced an autonomous AI action producing an unintended or problematic outcome.
The systems being governed are also changing after they have been approved. More than nine in ten organisations have stretched an AI system beyond its original purpose in the past year, meaning the AI in use today may be quite different from the system that was originally reviewed.
Against this backdrop, 99% of respondents say they would adopt a way to trace AI actions back to a responsible human identity. That question of identity is already becoming part of the governance debate. In June 2026, China issued a national standard giving every AI agent a verifiable identity, an early sign of where agent governance may be heading.
The report also breaks down where every APAC market and sector stands, and sets out what a provable approach to AI governance requires. Regulated, high-stakes work such as payments and anti-money laundering shows the most caution, precisely because a decision there has to be defensible to a regulator after the fact.
Read the full APAC State of Digital Trust: AI Governance Benchmark
Copyright © 2026 RegTech Analyst
Copyright © 2026 RegTech Analyst
