The Fintech Log

Bank AI adoption outpaces safeguards as SAS study finds widespread gaps in model oversight

Nov. 10, 2026  •  Article  •  AI in Fintech

Summary

Banks have moved AI into core risk and treasury work faster than they have established controls to oversee it, according to research published by SAS, FT Longitude and Microsoft on October 22, 2026. The survey of 300 risk, credit and treasury decision-makers across 31 countries found that 59% of responding banks had moved beyond early experimentation. AI was partially or widely embedded in 63% of core credit risk workflows and 55% of asset-liability management and balance-sheet risk steering processes.

Governance has not kept pace. Only 19% of surveyed institutions reported real-time monitoring for model drift, while 42% lacked documented, operational kill switches for autonomous models. Another 26% had no dedicated AI model risk policy. Although 76% expect AI to make autonomous credit and risk decisions at scale by 2031, only 10% say it does so today. The findings distinguish widespread AI use from the more demanding task of supervising autonomous decisions.

Full Article

Banks are embedding artificial intelligence in credit risk and treasury operations while leaving significant gaps in the controls needed to supervise it, according to a study published on October 22, 2026, by SAS, FT Longitude and Microsoft. The findings show that AI use is widespread, but real-time model monitoring, risk policies and mechanisms for stopping autonomous systems are far less established.

The report, Why AI at Scale Needs Governance at Speed, draws on a June–July 2026 survey of 300 risk, credit and treasury decision-makers in 31 countries. When traditional machine learning is counted alongside generative and agentic AI, 299 of the 300 banks represented in the survey use AI. That broad measure of adoption should not be confused with AI independently making decisions at scale: only 10% of institutions say it currently does so.

AI moves into core banking decisions

More than half of responding banks, 59%, said they had advanced beyond early AI experimentation. That includes 44% reporting partial embedding and 15% reporting widespread embedding of AI in core decision pipelines.

A functional breakdown of the research, highlighted in trade reporting on November 9, found AI partially or widely embedded in 63% of core credit risk workflows. The comparable figure was 55% for asset-liability management and balance-sheet risk steering, the processes banks use to manage the relationship between assets, liabilities and financial risks.

Those figures describe a range of AI uses, not solely agentic systems. The distinction matters because an AI tool used to support a decision presents a different oversight challenge from one authorized to act autonomously. The study found a large gap between current and anticipated autonomy: 76% of banks expect AI to make autonomous credit and risk decisions at scale by 2031, including 25% that expect this within two years.

The survey also points to tension between performance and explainability. Nearly two-thirds of senior banking executives, 63%, said they would deploy an opaque, or “black-box,” AI model if it offered better financial performance, returns or liquidity than an explainable alternative. The finding reports executives’ stated willingness, not evidence that those models have been deployed.

Monitoring and intervention controls lag

Only 19% of surveyed banking institutions run real-time monitoring to detect model drift, according to the research. Drift occurs when a deployed model’s behaviour changes relative to the conditions under which it was initially validated. Without real-time monitoring, a bank has less immediate visibility into whether a model continues to perform as expected.

Other safeguards are also uneven. The study found that 42% of institutions lack documented and operationalized kill-switch mechanisms to contain or deactivate autonomous AI models that exceed preset risk limits. More than one in four banks, 26%, reported having no dedicated AI model risk policy. Only 36% continuously retrain or refresh their production AI models.

Large language models pose a further oversight challenge. Some 92% of surveyed institutions acknowledged that at least one such model in their organization had been built or was operating entirely outside formal governance oversight. That figure refers to the presence of at least one ungoverned model at an institution; it does not mean that 92% of all models operate without oversight.

Taken together, the results identify separate tasks for banks: knowing which models they operate, checking whether those models remain within validated parameters and retaining the ability to intervene when an autonomous system breaches a limit. The survey does not establish that every institution with a control gap has suffered a model failure.

Governance becomes the test of scale

Marilin Pikaro, director of innovation, conduct and consumers at the European Banking Authority, cautioned that banks’ AI ambitions must match their governance capacity. Her warning addresses the pace of adoption relative to the controls available to supervise it.

Anselmo Marmonti, vice president of risk, fraud and compliance solutions at SAS, said failures by unmonitored AI decision agents can have consequences for customers, banks’ reputations and balance sheets. His comments underline why the shift from AI-assisted work to autonomous decisions raises the stakes for oversight.

The study was released on October 22; November 9 reflects later trade coverage, not a new report launch. SAS also said it planned to present the full findings and risk-governance frameworks at RiskMinds International 2026 in London, scheduled for November 16–19. For banks planning to give AI greater decision-making authority, the research sets out a practical question: whether monitoring, policies and intervention controls will be ready when that authority expands.


By 'TFL Editorial Team'
Researched with AI tools. Edited and fact-checked by our editors.

References


You May Also Like

Banks have moved AI into core risk and treasury work faster than they have established controls to oversee it, according to research published by SAS, FT Longitude and Microsoft on October 22, 2026. The survey of 300 risk, credit and treasury decision-makers across 31 countries found that 59% of responding banks had moved beyond early experimentation. AI was partially or widely embedded in 63% of core credit risk workflows and 55% of asset-liability management and balance-sheet risk steering processes. Governance has not kept pace. Only 19% of surveyed institutions reported real-time monitoring for model drift, while 42% lacked documented, operational kill switches for autonomous models. Another 26% had no dedicated AI model risk policy. Although 76% expect AI to make autonomous credit and risk decisions at scale by 2031, only 10% say it does so today. The findings distinguish widespread AI use from the more demanding task of supervising autonomous decisions.

Nov. 10, 2026 • 3 min read

Stripe’s growing role in AI-assisted shopping has raised concerns at Visa and Mastercard about how payments are routed and what information reaches banks and card networks. At merchants outside Stripe’s checkout network, its Link wallet can give an AI agent a capped virtual card to complete a purchase, then charge the shopper’s saved payment method separately. That structure can leave the shopper’s issuing bank without a direct view of the merchant. The distinction matters as shopping gains traction inside Meta’s Muse agent and Stripe reports sharp growth in agentic checkout volume. Stripe says it developed the virtual-card feature in consultation with network partners. Visa and Mastercard are advancing a different approach, using tokenized credentials linked to an existing card and controls designed to verify agents and spending permissions. Mastercard says multiple methods may coexist, while banks have called for auditable credentials that preserve the connection between consumers and merchants.

Nov. 9, 2026 • 3 min read

Visa’s Agentic Commerce Benchmark found that 22 AI models made suitable shopping decisions in 57.2% of test runs, leaving a 42.8% failure rate. The assessment tested whether agents could identify products that met a buyer’s requirements—or decline to buy when none did. Models that gathered more buyer details through conversation performed better, but even the strongest model fell short of a perfect score. The findings concern shopping decisions, not a direct test of payment execution. The results add to questions about how autonomous purchasing could operate at scale. Visa’s consumer research found limited trust in AI systems making payments, while its survey of acquirers identified liability and compliance concerns. Disputed agent transactions currently fall under existing card-not-present rules unless an authenticated, tokenized flow shifts liability. Payment-industry work on agent identity and standards is advancing, but a unified approach to disputes over non-fraud purchases by autonomous agents has not been established.

Nov. 7, 2026 • 3 min read

Mastercard is seeking to make network tokens the payment standard for purchases initiated by AI agents. On its October 29, 2026, earnings call, Chief Executive Michael Miebach described processor-issued one-time virtual card numbers as a temporary bridge rather than a lasting approach. He said Mastercard would support end-to-end tokenization through Agent Pay to preserve cardholder choice, rewards and spending controls. The company also announced that Shopify is integrating Agent Pay and network tokens into Shop Pay, alongside a strategic partnership with OpenAI covering AI payments and services. Mastercard has been building merchant-facing tools for agentic commerce throughout 2026, but Miebach acknowledged that merchants’ technical readiness remains uneven. Other payments companies are developing their own approaches: Stripe supports agentic network tokens from Mastercard and Visa, while Visa has introduced tools for AI-driven shopping. Consumer concerns about fraud and liability remain a potential barrier to adoption.

Oct. 29, 2026 • 3 min read

The Linux Foundation appointed payments veteran Michael Hursta executive director of the x402 Foundation on October 1, 2026, as payment systems designed for autonomous AI agents gained attention. Foundation members reported more than 75 million agentic payment transactions in a 30-day period in September, and Coinbase and Amazon Web Services showcased production payment rails at Sibos 2026. Reported transaction counts, however, do not establish the value or commercial character of that activity. The emerging protocols address different obstacles. Coinbase’s x402 enables per-request USDC payments through HTTP, while Google’s Agent Payments Protocol uses signed mandates to connect an agent’s purchases to a user’s authorization. Stripe and Tempo have also proposed session-based billing through the Machine Payments Protocol. These approaches respond to card fees and interactive checkout checks that can impede small, automated transactions. Card networks are developing delegated-identity tools, leaving multiple potential paths for agent payments rather than a single replacement for existing rails.

Oct. 28, 2026 • 3 min read

Mastercard opened developer early access for Agent Pay for Machines (AP4M) on October 20, 2026, releasing buyer- and seller-side APIs, a sandbox and developer toolkits. In a Cloudflare Connect demonstration, marketing agency Hype tested an autonomous agent making nine payments to two merchants within a single budget, without manual billing or account setup. AP4M uses HTTP 402 payment challenges, but differs from crypto-native implementations of the x402 protocol. Its agents issue digitally signed vouchers off-chain; merchants can batch and redeem them later instead of settling every request on-chain. The pilot combined agent wallet infrastructure from Coinbase Developer Platform and BVNK, settlement on Base, and Mastercard’s multi-currency fiat and stablecoin settlement platform. Circle separately expanded its managed payments service to target AI-agent traffic using USDC. Together, the developments show competing approaches to machine-to-machine payments, with different choices about settlement, funding and who manages digital assets.

Oct. 27, 2026 • 3 min read