AI Accountability Outpaces Evidence at Singapore Firms
Source: The Edge Singapore
Singapore companies are racing to deploy AI across their operations, but most cannot prove how their systems reach decisions — a gap that threatens to undermine the trust frameworks regulators and boards have spent years building. As the MAS rolls out agentic AI safeguards and companies cheerfully assign...
Singapore companies are racing to deploy AI across their operations, but most cannot prove how their systems reach decisions — a gap that threatens to undermine the trust frameworks regulators and boards have spent years building. As the MAS rolls out agentic AI safeguards and companies cheerfully assign accountability to specific teams, the hard evidence that makes governance meaningful is largely missing.
A new study by Sumsub and the Singapore FinTech Association finds that while 70% of Singapore businesses have designated a person or team responsible for AI outcomes, only 29% can produce an audit trail showing how an AI-driven decision was reached. "They own the destination, but they can't prove the route," said Penny Chai, Sumsub's vice president for Asia Pacific. The top barriers: complexity of AI models (66%), integration difficulties across technology platforms (50%), and trouble tracking actions by third-party AI tools (49%).
The findings land at a pivotal moment. The Monetary Authority of Singapore recently introduced its Safeguards for Agentic Finance at Runtime (SAFR) framework — one of the world's first operational guardrails for AI agents in financial services — while the government's Model AI Governance Framework for Agentic AI pushes beyond paper-based policies toward practical, verifiable controls. The study, which surveyed 720 senior professionals across nine APAC markets, gave Singapore an overall AI governance score of 65.6, slightly below the regional average of 67.1. Chai noted that this likely reflects Singapore firms grading themselves against stricter local standards rather than lagging in implementation.
In practice, companies are keeping AI on a short leash. Some 94% of Singapore respondents reported using or piloting multi-step AI systems, and 90% expressed confidence in letting AI handle lower-risk tasks. But only 16% said they had significantly increased the scope or autonomy of their systems over the past year. Holly Fang, president of the Singapore FinTech Association, pointed out that financial institutions weigh monetary loss, reputational damage, and their licence to operate when evaluating AI systems — adoption has stayed slowest in areas touching payments or direct money movement. The liability question cuts both ways: AI vendors selling into financial services are increasingly fielding demands about who bears responsibility when automated decisions go wrong.
Why it matters for Singapore: The audit trail gap is not just a compliance footnote — it is the hard problem underneath every governance framework Singapore has built. If companies cannot reconstruct how their AI reached a decision, frameworks like SAFR and the Model AI Governance Framework remain aspirational. Closing that gap will determine whether Singapore's regulatory leadership in AI governance translates into systems that are genuinely accountable, or whether the paperwork outruns the plumbing.
