99% of Firms Want to Trace AI Decisions Back to a Human, But Are They Ready F...
Source: Fintech News SG
Singapore's banks and fintechs have been quick to put AI agents to work, but the industry is confronting a question beyond model performance: can anyone prove who authorised an AI decision? Sumsub's APAC benchmark finds 98.6% of firms want traceability, yet only 38% keep a tamper-proof trail.
Singapore's banks and fintechs have been quick to put AI agents to work — approving transactions, screening compliance alerts, answering customers — but the industry is now confronting a question with nothing to do with model performance: when an agent acts, can anyone prove who authorised it and why? That gap between using AI and accounting for it is the central finding of Sumsub's new APAC State of Digital Trust: AI Governance Benchmark, a survey of 720 senior technology, risk and compliance professionals across nine Asia-Pacific markets.
The report finds near-universal demand for accountability — 98.6% of organisations say they are likely to adopt software that traces AI actions back to a responsible human identity. Yet readiness lags badly. Across every market and sector surveyed, Traceability scored 61.0 out of 100, trailing Autonomy and Responsibility at roughly 70 each. Sumsub calls the gap the "Accountability Asymmetry": while 95% of respondents are confident they can explain an AI decision, only half can reconstruct the pathway that led to it, and just 38% keep a tamper-proof audit trail. The top worry, cited by 52%, is that AI produces incorrect or unreliable decisions no human can yet defend.
The financial sector, unsurprisingly, is furthest along — and Singapore's own banks are part of that picture. Financial services firms posted the highest overall governance score at 69.6, with 72% already running multi-step AI systems in production and 68% keeping audit trails of AI decisions. CIMB Singapore's digital and AI adoption lead Edmund Phuang points to one reason the sector is investing in governance: large language models can institutionalise knowledge, letting new joiners build on a bank's accumulated best practices instead of starting from scratch. The report lands as the Monetary Authority of Singapore prepares to finalise its AI risk guidelines covering agentic AI — a reminder that the regulator is watching exactly the autonomy-versus-proof dynamic the survey measures.
The commercial stakes are becoming clearer too. Sumsub's APAC vice president Penny Chai frames traceability as the prerequisite for scaling high-stakes AI safely, and the report argues that verified accountability is turning into a competitive edge: counterparties are more willing to let an agent into their systems when it arrives with a verifiable identity and a clean record. In other words, the same evidence that satisfies a regulator doubles as a signal of trustworthiness in the market — an agent you cannot reconstruct is a liability you cannot price.
Why it matters for Singapore: Singapore has positioned itself as a hub where AI deployment and AI governance grow together, from MAS's agentic AI guidelines to the industry's own assurance frameworks. This benchmark suggests the next differentiator for local banks, fintechs and enterprises will not be how much AI they run, but how well they can prove what it did. The organisations that close the traceability gap first — with immutable audit trails and reconstructable decision pathways — will be the ones regulators clear to scale and partners trust to transact.
