MAS tells financial institutions to begin quantum preparation five to ten years before arrival

The ordering is inverted from how capital planning usually runs. Quantum computing is the furthest away at five to ten years ^[4], and it is the item Chia told institutions to begin preparing for now ^[5]. Tokenisation, which he put several more years from scale ^[3], got no comparable instruction. Five to ten years from Friday's address puts the machines somewhere between 2031 and 2036 ^[1].

Chia's sixfold rise puts the preceding three-year average near 367 high-severity CVEs a year, which makes this year's 2,200 an increase of roughly 1,833 ^[2]. He also cited CrowdStrike's reported 89% increase in AI-enabled cyber attacks ^[8].

The productivity side is thinner. Corporate adoption is accelerating while a much smaller proportion of companies report significant productivity gains ^[13]. Chia said those gains were likely to increase as employees and organisations become better users of AI through training, process redesign and new products ^[14]. MAS did not quantify either proportion.

The paper trail runs back to 2023. MAS published a generative AI risk framework with the industry that year, then two AI Risk Management Handbooks in 2025 covering banking, insurance and capital markets ^[9]. The framework came three years before the consultation now in progress ^[3]. Chia said the guidelines would establish "what" financial institutions should do, while the handbooks would provide guidance on "how" to implement those requirements ^[11]. SAFR arrived in July as a white paper, while the guidelines are still out for public consultation ^[12][10].

For large, well-managed institutions, MAS is focusing on safety, guardrails and accountability instead of encouraging adoption ^[15]. The encouragement goes to Pathfin.ai, the platform MAS launched to share and match validated AI solutions across the industry ^[17]. "We should avoid a winner-takes-all dynamic if we are to maintain a competitive and stable financial system to support the public and the economy," Chia said ^[16].

In my view the consultation moves money in the next twelve months and the quantum sentence does not, because a consultation has a response date and a five-to-ten-year estimate does not. The counter case is the 2023-to-2026 record: MAS wrote a framework, then handbooks, then supervisory expectations ^[9][10], so a line in a speech is how the sequence starts. Chia also said AI models already perform at expert or specialist levels in coding, graduate-level science and mathematics and general knowledge work, with gaps in complex interpretation, strategic judgment, decision-making and human interaction ^[20]. Those gaps are what keeps a human in the approval chain and the compliance headcount on the books.

I would be wrong if quantum inventory work turns up in this year's capex on the strength of an arrival date alone. The institutions most likely to do it are the ones already running AI at scale in fraud detection and credit underwriting ^[19].