Sumsub Survey Finds 96% of Indonesian Organisations Say They Can Explain AI Decisions, but Only 34% Keep Full Records

Indonesian organisations are handing more decisions to artificial intelligence than they can later prove, according to a regional survey by identity verification and compliance company Sumsub reported by Investor.id. The study found that 96 per cent of Indonesian organisations believe they can explain the decisions their AI systems make, yet only 34 per cent keep a complete, tamper-proof record of those decisions.

The findings come from the APAC State of Digital Trust: AI Governance Benchmark, an independent survey of 720 senior professionals in nine Asia-Pacific countries carried out by Blackbox Research with the Singapore FinTech Association. Sumsub calls the gap an ‘accountability asymmetry’: the distance between how much organisations rely on AI decisions and how well they can show what those decisions actually were.

Other results point the same way. Some 91 per cent of Indonesian respondents said they are comfortable letting AI make routine decisions without a human in the loop, but that falls to 55 per cent when financial transactions are involved. Sixty-four per cent said they had already seen AI act on its own in a way that caused a problem or an unwanted result, and 94 per cent had used an AI system beyond its original purpose in the past year. Data quality was the top concern for 56 per cent, the highest share in the region.

What it means in practice

The shift matters because Indonesian companies are moving from AI tools that help staff to agentic systems that plan, decide and act for the organisation. When an agent approves a payment or rejects an application, regulators and customers will expect a trail showing what it did and who was responsible. The National Research and Innovation Agency (BRIN) said the shift opens opportunities but demands adaptive and responsible development.

Regulatory pressure is building. The financial services authority, OJK, has already set minimum standards for how banks govern AI across its life cycle, from risk management to audit. Two presidential regulations, on the national AI roadmap and AI ethics, are awaiting the President’s signature, following drafts that would create a national coordination task force and three risk tiers. Bank Indonesia has also warned that AI could fuel herding behaviour in markets.

Sumsub, which sells compliance tools, has an interest in the problem it describes, and the survey is a snapshot of perceptions rather than an audit. Still, the gap between confidence and record-keeping is wide. Sumsub has set up an Agentic AI Council in the region to draft practical frameworks. What to watch next is whether OJK and other regulators begin asking companies to show decision logs, and how the presidential regulations treat accountability for autonomous systems.