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 links to alternate free versions.

NEWS:

Buy-Side AI Adoption Hindered by Broker Research Licensing Restrictions, Traders Magazine, July 16, 2026.

• Latest Substantive Research and Aiera survey finds that 77% of buy side firms currently have organisation-wide deployments of generative AI platforms in place (e.g., Claude, ChatGPT)
• 77% said that broker research was the most valuable input to receive as machine-readable feeds directly into their internal AI systems, followed by Earnings Transcripts (57%) and Market Data (42%).
• Broker/Data Licensing Restrictions are the biggest barrier preventing adoption of direct research/data feeds (69%), followed by Compliance and Entitlements (54%)

Substantive Research, the research and market data discovery and pricing analytics provider, has partnered with Aiera, a consortium-backed content delivery and access platform, to better understand how the buy side is building an effective and sustainable AI infrastructure to remain competitive. Jointly, they surveyed 35 of the largest global asset managers and here are the key findings:

• 77% of buy side firms currently have organisation-wide deployments of generative AI platforms in place (e.g., Claude, ChatGPT etc.)
• Broker/Data Licensing Restrictions are the biggest barrier preventing adoption of direct research/data feeds (69%), followed by Compliance and Entitlements (54%)
• 37% said that approving, adopting and onboarding these models takes 4-6 months, for 20% it took over 6 months, and for 17% it only took 1-3 months
• Just over a quarter are evaluating or have implemented vertically integrated AI platforms that specialise in finance/investment research
• 44% view specialised platforms as potentially strategic long-term partners, with the same number evaluating but undecided
• 77% said that broker research was the most valuable input to receive as machine-readable feeds directly into their internal AI systems, followed by Earnings Transcripts (57%) and Market Data (42%)

Why AI Needs Light, TabbForum, July 9, 2026: Modern AI is no longer a single processor doing the work. It is hundreds of thousands of chips that have to operate as one system, passing enormous volumes of data back and forth constantly. And the electrical connections the industry has relied on for decades are running into hard physical limits on how much data they can carry, how far, and at what energy cost. Thomas DiFazio, an ETF Strategist at Roundhill, explains how the answer is light. For two years, the AI hardware trade has been about raw power. Faster chips. More memory to keep them fed. And, the market has rewarded companies supplying both legs. There is a third piece that has gotten far less attention, and it is becoming just as important: moving data between the chips. Modern AI is no longer a single processor doing the work. It is hundreds of thousands of chips that have to operate as one system, passing enormous volumes of data back and forth constantly. As these systems scale, the connections between chips increasingly determine how well the whole thing performs. And the electrical connections the industry has relied on for decades are running into hard physical limits on how much data they can carry, how far, and at what energy cost. The answer is light. Replacing electrical signals with optical ones is the basis of one of the most talked-about and least understood themes in the market: photonics and optics. Modern AI is no longer a single processor doing the work. It is hundreds of thousands of chips that have to operate as one system, passing enormous volumes of data back and forth constantly. And the electrical connections the industry has relied on for decades are running into hard physical limits on how much data they can carry, how far, and at what energy cost. Thomas DiFazio, an ETF Strategist at Roundhill, explains how the answer is light.

Revolut Is Building An AI Brain For Banking, And It Could Change Finance Forever, Forbes, Jul 8, 2026. Revolut is pioneering a unique AI strategy in banking with its proprietary foundation model, PRAGMA, designed to understand financial behavior holistically. Unlike competitors using fragmented AI tools, PRAGMA integrates all customer interactions—transactions, app usage, investments, and support requests—into a single connected system. This unified approach significantly boosts fraud detection by 64.7%, enhances credit risk prediction by 16%, and improves product recommendations by 41%. Powered by NVIDIA GPUs, PRAGMA scales to serve 70 million users, allowing insights from one function to improve performance across the entire customer journey. This strategy fosters organizational agility, enabling personalized services and advanced agentic AI, setting a new standard for enterprise AI in finance.

JPMorgan Builds AI Agents That Beat 60/40 Portfolio in Backtests. Bloomberg – no paywall, July 9, 2026. Wall Street banks have spent the past two years embedding large language models into research, coding and internal investing tools, exploring whether those systems can move beyond assisting workers to making one of the industry’s most consequential decisions: how to allocate capital across markets.

• JPMorgan has been testing an array of AI-powered investing agents that shift between stocks and bonds depending on changing market conditions. And the early results are encouraging.
• Yet while a growing body of evidence suggests that widespread adoption of AI may make investors faster and more informed, it also raises questions about the market-wide consequences if everyone turns to these models.
• Moreover, a recent Harvard-led study found that a machine-learning algorithm can predict about 71% of mutual-fund trading decisions, but the trades it failed to anticipate appeared to be where most of the value lies.
• All that said, the advances are leveling the playing field for fund managers.

PAPERS:

NBER

Assessing the Benefits of Optimized Agentic AI Systems for Asset Pricing. Ralph S. J. Koijen & Bradford Levy. Working Paper 35431. DOI 10.3386/w35431. Issue Date July 2026

Evaluating optimized AI systems for asset pricing is fundamentally difficult for two reasons. First, models are trained on all data, implying that any backtest or analysis using historical data suffers from look-ahead bias. In addition, markets are reflexive — as investors adopt AI, prices adjust — which may erode the very patterns the AI system was trained to exploit. We introduce a real-time, out-of-sample benchmark designed to sidestep both problems. The benchmark measures how well AI systems can explain contemporaneous stock returns around earnings announcements using only information available at announcement time, including the text of the announcement itself. Applying this benchmark to a range of agentic AI systems — which extract structured signals from earnings call transcripts and optimize over those signals — we find that the best-optimized systems more than double the explained variation in returns relative to standard benchmarks (R2 increasing from 8% to close to 20%). We show that AI-based optimization can deliver efficiency gains relative to traditional machine learning methods while also improving interpretability as our approach produces human-readable economic mechanisms that explain price movements. These learned rules can be compared to the drivers of realized returns in existing asset pricing models to identify missing sources of variation in a data-driven, self-evolving way that integrates empirical learning with economic structure. We release an SDK for researchers to improve on our results. Saturating this benchmark would represent fundamental progress in understanding how capital markets process firm-level information.

Risk Design: AI and Prediction Beyond Screening in Insurance Markets. Alex Chan. Working Paper 35444. DOI 10.3386/w35444. Issue Date July 2026

I study insurance markets in which scalable prediction, like AI, designs residual risk rather than merely classifies fixed risk. A complete-contracting benchmark shows that if prevention is observable, contractible, competitively supplied, and fully priced, it does not matter whether consumers, insurers, or vendors supply it. Adverse selection breaks such irrelevance. When high-risk consumers are more “AI-treatable,” efficient prevention makes low-risk contracts attractive to them. A contract intended for low-risk consumers faces a risk-design trilemma: separate, prevent efficiently, or avoid cross-subsidy, but not all three. The result extends Rothschild-Stiglitz from distorted coverage to distorted risk-control technology and offers market design insights of AI in insurance markets.

AI Premium. Nicola Borri, Aleh Tsyvinski & Yukun Liu. Working Paper 35451. DOI 10.3386/w35451. Issue Date July 2026

Using 380 trillion tokens of realized AI consumption across more than four hundred large language models from the licensed proprietary OpenRouter dataset covering approximately 2 percent of current global monthly AI token consumption, we analyze how AI affects firms, markets, and workers. Leveraging the unprecedented size, scope and granularity of this data, we construct the AI Factor from growth in tokens, dollars, and users, estimate firm-level AI Betas from stock return comovement, and characterize the AI Premium. First, we build a high-frequency AI factor and decompose it into salient components. Second, we show that firms whose returns covary more positively with the AI factor—high AI beta firms—earn higher subsequent returns, and the AI premium is large and heterogeneous. A value-weighted longshort strategy earns 64.1 basis points per week, and the premium is large for loadings on the intensive, frontier-oriented margin of AI consumption—closed-source models, paying and seasoned users, and long prompts—but not on casual or open-weight use. Third, the premium reaches beyond technology firms into consumer-facing and capital-heavy parts of the economy, but is absent in emerging markets, including China. Fourth, the AI exposure is more positive in nonroutine interactive work and more negative in analytical, scientific, and operations-control skills—an occupation one standard deviation higher in interaction-and-communication content has 0.36-standard-deviation higher market-implied AI exposure. Additionally, we provide early evidence of the rise of the agentic economy.

NOGs/IGOs

Bank for International Settlement – BIS

The AI investment race. by Phurichai Rungcharoenkitkul. BIS Working Papers |  No 1367  | 14 July 2026