MAS warns of risks in AI-driven digital-asset trading

SINGAPORE, 10 OCTOBER 2026 —

Singapore’s central bank warns AI-driven trading could destabilise markets as agents take over transactions

The Monetary Authority of Singapore (MAS) has flagged growing risks in financial markets as artificial intelligence agents move beyond analysis to autonomously executing trades, potentially accelerating capital flows and undermining liquidity controls without proper safeguards. Speaking at an INSEAD conference on October 9, MAS Managing Director Chia Der Jiun cautioned that while AI agents promise efficiency and accessibility, their unchecked deployment could disrupt market stability—particularly in digital-asset trading—unless governance frameworks are urgently strengthened.

The warning comes as blockchain and tokenised assets transition from experimental phases to commercial adoption, with MAS emphasising that trust and regulatory clarity are now the critical barriers to scaling these technologies. Chia’s remarks underscore a broader regional debate over how to balance innovation with systemic risk, as AI-driven trading agents—already operational in some institutions—could reshape market dynamics by prioritising algorithmic optimisation over human oversight.

Chia’s address at INSEAD’s Future of Digital Finance conference, themed around infrastructure, AI, and regulation, highlighted three immediate concerns: the speed of capital movement, the erosion of liquidity buffers, and the lack of accountability when AI agents act independently. “We are entering a world where AI agents may increasingly act on behalf of individuals and institutions, not only to analyse information but also to initiate and execute financial transactions,” he said.

“This could result in money moving around faster—for smaller differences in benefits and incentives—while financial institutions struggle to sharpen their value proposition in response.”

The MAS official stressed that autonomous trading systems, if left unchecked, could lead to “significant implications” for collateral management and risk limits. “While agents create opportunities for efficiency, accessibility, and innovation, they must be deployed safely,” Chia stated.

“This requires controls around authority, permissions, identity, accountability, and governance.” He added that institutions must ensure AI agents adhere to user instructions, internal policies, and regulatory mandates—an increasingly complex task as agents operate across jurisdictions and asset classes.

MAS has already taken preliminary steps to address these risks. In early 2026, it published the Safeguards for Agentic Finance at Runtime (SAFR) white paper in collaboration with industry partners, outlining principles for real-time oversight of AI-driven financial systems. The authority has also released an open-source reference implementation of SAFR through its Future of Finance Institute, encouraging experimentation while building a framework for broader adoption.

“The technology is still developing rapidly, and many questions remain,” Chia acknowledged. “A public-private partnership approach will remain essential to form coalitions of expertise and establish standards.”

The call for collaborative regulation was echoed by Norberto de Andrade, senior adviser for AI policy at the Digital Currencies Governance Group, who noted that traditional legislative approaches are being supplemented by “soft law” instruments such as codes of conduct and regulatory sandboxes. “We see more ways to regulate than just top-down legislation,” Andrade said during a panel discussion on AI agents and blockchain infrastructure.

“This is promising for policymakers working on emerging technologies, as regulations must address not just the tools but their effects, behaviours, and consequences.”

Marco Dal Lago, chief expansion officer at stablecoin issuer Tether, reinforced the need for targeted regulation. “If we understand why we are regulating, then we can determine how and when to do so,” he said. The distinction reflects a growing trend in Asia, where regulators are adopting agile, technology-specific frameworks rather than rigid, one-size-fits-all policies.

For Malaysia and regional markets, MAS’s warnings carry particular weight given Singapore’s role as a hub for digital-asset innovation and cross-border capital flows. The Bank Negara Malaysia (BNM) has previously signalled cautious optimism about AI in finance, including a 2025 discussion paper on Responsible AI in Financial Services, but has not yet addressed the specific risks of autonomous trading agents.

If MAS’s concerns materialise, Malaysian institutions—particularly those engaged in digital-asset trading or tokenisation projects—may face pressure to adopt similar safeguards to prevent liquidity shocks or regulatory arbitrage.

The immediate next step for MAS will be refining its SAFR framework in consultation with industry stakeholders, with Chia indicating that further public-private collaborations will be critical. Meanwhile, the INSEAD conference’s discussions suggest that the evolution of AI regulation will hinge on balancing speed with stability—a challenge that will test the resilience of both markets and policymakers in the coming years.