This semi-monthly column highlights news, government documents, NGO/IGO papers, conferences, industry white papers and reports, academic papers and speeches, and central bank actions on the subject of AI’s fast paced impact on the banking and finance sectors. The chronological links provided are to the primary sources, and as available, indicate alternate links with no paywall, as available.
NEWS:
OpenAI gives Japan banks access to latest model, Japan’s finance minister says. Reuters, May 29, 2026. “OpenAI has given some Japanese financial institutions access to its GPT-5.5 model to help prevent cyberattacks, Japanese Finance Minister Satsuki Katayama said on Friday after a meeting with the U.S. company’s chief strategy officer.”
Real-Time Risk: How AI Changes the Speed of BankingBanking risk management was designed for a batch-processing world. AI is enabling real-time risk assessment — and the banks making this transition are seeing risks that batch systems miss entirely.
Plaid Launches New AI Data Tools, Guaranteed Payments to Overhaul ACH. This Week in Fintech, May 21, 2026. Financial infrastructure provider Plaid** introduced two in-house foundational artificial intelligence models on Thursday that aim to dramatically improve how the company classifies transaction data, capping off a week of back-to-back security and payment launches ahead of its annual Plaid Effects conference. The rollout follows the May 20 launch of Guaranteed Payments, a service where Plaid assumes full financial liability for slow-moving bank transfers to eliminate multi-day settlement delays. By using real-time data from its network – which monitors over 500 million accounts – Plaid approves Automated Clearing House (ACH) transfers instantly, allowing digital applications to credit user accounts immediately. If an approved payment later fails, Plaid steps in to cover the financial loss. “If you’re using this guaranteed ACH product, and we’ve said we stand behind this transfer, then we take the liability on for those funds arriving,” Will Robinson, Plaid’s Chief Technology Officer (CTO) told This Week in Fintech in an interview. The launches build on a pipeline of product updates that the fintech company has deployed over the last month to modernize bank-level infrastructure. Below is a summary of all the recent infrastructure updates.
I’m the CEO of Goldman Sachs. The AI Job Apocalypse Is Overblown. New York Times Opinion, May 21, 2026. AI will absolutely disrupt the job market, but the US has a long track record of creating new jobs in response to disruption. The growing demand for data centers has created more than 200,000 construction jobs since 2022. AI may eliminate jobs in some sectors, but it will lead to growth in others. The US economy can and will adapt to major advances in technology.
What Real-Time Risk Looks Like. AI enables risk assessment at the speed of the business. LinkedIn. May 25, 2026.
- Real-time credit exposure: As transactions execute, credit exposure to counterparties is recalculated continuously. Limit breaches are detected in real time, not discovered the next morning.
- Intraday market risk: AI monitors trading portfolio risk metrics continuously, flagging when risk parameters approach limits. Traders receive alerts before limits are breached, not after.
- Real-time fraud scoring: Every transaction scored in under 50 milliseconds against 200+ behavioural features. Fraudulent transactions blocked at the point of authorisation.
- Dynamic liquidity monitoring: Liquidity positions updated continuously as payments flow. Predictive models forecast intraday liquidity needs, enabling proactive buffer management.
- Operational risk sensing: AI monitors system performance, process exceptions, and operational indicators in real time, identifying emerging operational risks before they become incidents.
AI set to automate up to 50% of tasks in most financial services roles. A UK government-commissioned report warns that AI could automate 30% to 50% of tasks across most financial services jobs, reshaping workforce planning, hiring, and required skill sets over the next decade. The report predicts growing demand for data, governance, software engineering, product design, and critical thinking skills, while also warning that entry-level career paths may shrink as agentic AI systems take on more operational work. Financial industry leaders, including executives from HSBC, Standard Chartered, and JPMorgan, are already openly acknowledging that AI will reduce headcount in certain banking roles even as firms hire more AI-focused talent.
Inside Claude’s rapid expansion across corporate finance. CFO, May 21, 2026. Big Four firms, Wall Street banks and finance teams are continuing to embed Anthropic’s AI platform into forecasting, reporting and day-to-day finance workflows.. Claude is becoming deeply embedded across corporate finance workflows, with Anthropic launching finance-focused agents for reconciliations, valuation reviews, earnings analysis, and statement audits, plus integrations across Microsoft Office. The piece highlights growing adoption by firms like PwC, KPMG, JPMorgan, Goldman Sachs, Citi, AIG, and Visa, while noting that many of Anthropic’s strongest use cases are concentrated in institutional finance workflows where enterprise data connectors are already in place.
Understanding the modern cybercrime landscape. It’s time for a re-think of the network’s pivotal role and how it can manage an enterprise’s digital defenses. MIT Technology Review. May 19, 2026. Financial pressures – The first factor arguably contradicts its neighbor in the landscape: general financial constraints and the pressure on CISOs and CIOs to achieve more with less. Despite the strategic reliance on the network and the expectation that it will be protected from cyber threats regardless, the appropriate latticework of defenses (e.g., skilled and right-sized IT teams using progressive tools and meaningful data insights, plus constant workforce education) is not always properly funded and sustained, particularly in the current tough economic climate.
Vendor Lock-In and AI: The Risk Banks Aren’t Pricing LinkedIn, The Lock-In Problem – AI vendor selection in banking typically evaluates features, price, and integration capability. These are necessary. They’re insufficient. The question most procurement processes don’t ask: what happens if we need to switch? AI systems are inherently sticky. The model is trained on your data. The integration is built into your workflows. Your teams have learned the platform. The switching cost is significant and grows with every month of operation. In a market heading for vendor consolidation — where some vendors will be acquired, some will pivot, and some will fail — this stickiness becomes a strategic risk. Why AI Lock-In Is Different – Traditional software lock-in is about data and workflow migration. AI lock-in includes additional dimensions:
- → Model dependency: Your AI models are trained on patterns specific to your data within the vendor’s architecture. Migrating to a new platform means retraining, which means months of performance regression.
- → Integration complexity: AI systems connect to core banking, CRM, fraud management, and compliance systems. Each integration point is a migration cost.
- → Institutional knowledge: Your teams understand the vendor’s tools. Knowledge transfer to a new platform takes 6–12 months.
- → Regulatory documentation: Model validation documents, regulatory approvals, and audit trails are tied to specific implementations. Migration requires re-validation.
PAPERS:
NBER
The Optimal Use of AI in Financial Regulation. Working Paper 35227. DOI 10.3386/w35227. Issue Date
We study whether AI methods applied to large-scale portfolio holdings data can improve macroprudential financial regulation. We build a graph-based deep learning model tailored to security-level data on the holdings of financial intermediaries. The architecture incorporates economic priors and learns latent representations of both assets and investors from the network structure of portfolio positions. Applied to the universe of non-bank financial intermediaries, covering nearly $40 trillion in wealth, the model substantially outperforms existing approaches in out-of-sample forecasts of intermediary trading behavior, including in crisis episodes. The model has more than ten times the explanatory power for the cross-sectional variation in asset returns during stress events compared to traditional approaches, and it outperforms existing systemic risk metrics at the institution level. Its learned representations show that the holdings network encodes rich, economically interpretable information about fire- sale vulnerability. The architecture is fully inductive, producing informative estimates even when entire asset classes or investors are withheld from training. We embed our empirical approach into a macroprudential optimal policy framework to formalize why these objects matter for policy and welfare. We show that even in an equilibrium environment subject to the Lucas critique, the predictive information from the model improves welfare by sharpening the cross-sectional targeting of policy interventions, and we demonstrate a complementarity between prediction and structural knowledge.
Preference for Explainable AI. Working Paper 35240. DOI 10.3386/w35240. Issue Date Participants acted as loan officers deciding whether to approve real $10,000-loans issued by a private U.S. lender using an AI’s default-risk predictions. When explanations revealed that the AI penalized non-White or female borrowers, participants were more likely to override the AI’s profit-maximizing recommendation. When their bonuses depended on repayment, however, they sought predictions but avoided explanations, consistent with willful ignorance; this effect faded when explanations were framed as purely financial or demographics were hidden. A secondary experiment reveals a novel bias: participants failed to reason contingently and undervalued explanations even when these complemented private information and improved decision accuracy.
Central Banks:
European Central Bank
Financial Stability Review, May 2026 – The Financial Stability Review provides an overview of potential risks to financial stability in the euro area. It aims to promote awareness in the financial industry and among the public of euro area financial stability issues. It is published twice a year, with the next release set for November 2026…
In addition, several cross-cutting structural challenges remain critical for financial stability, with the potential to amplify existing cyclical vulnerabilities. These include rising vulnerabilities associated with cybersecurity weaknesses and hybrid threats against critical infrastructure in an increasingly complex geopolitical landscape. Moreover, the rise of artificial intelligence, especially newly emerging frontier models, and quantum computing provide not only opportunities but also risks of destabilisation along the innovation path. Additionally, risks stemming from global regulatory fragmentation and deregulation, challenges linked to ageing populations and rising risks associated with climate change, including the materialisation of physical risks, remain significant concerns. The potential for these cyclical and structural vulnerabilities to crystalise simultaneously and amplify each other heightens the materiality of risks to euro area financial stability.
NGOs/IGOs
Bank for International Settlement
The geography of AI firms. BIS Working Papers | No 1343 | 20 April 2026
Artificial intelligence (AI) is advancing rapidly, yet relatively little is documented about the global distribution of AI production. This gap is especially important for policymakers, as many economies are evaluating their strategic priorities in AI production and some are considering sovereign AI strategies. We map AI firms globally using novel data and large language model-based textual analysis.