BankingNewsAI Daily Brief ·
FSB Chair sets out work on AI and stablecoins
Banking AI
Financial institutions & fintech technology
FSB Chair sets out work on AI and stablecoins
For the chief risk officer at a US bank
The FSB has made responsible AI a financial-stability and governance issue, not just an internal control task. You need a control map that can stand up to new expectations for governance and information sharing.
The Financial Stability Board published Chair Andrew Bailey’s letter to G20 finance ministers and central bank governors, which introduces finalised sound practices for responsible AI adoption. The letter introduces reports being delivered at the request of the US G20 Presidency, ahead of the G20 meeting on 15 October. Bailey cited AI risks alongside volatility in sovereign bond markets and vulnerabilities in private credit markets, and called for continued vigilance to safeguard financial stability. The FSB said safe AI adoption across the financial system will remain an important part of its agenda next year. The stablecoin note outlines further work to enhance monitoring of cross-border vulnerabilities and foster cooperation and information sharing.
→ Action
Model risk: Inventory AI uses and map governance controls to the FSB’s final responsible-AI practices.
Full story →Read article → from Fsb
Nubank expands in US, deploys AI tools, rules out Monzo deal
For the fintech partnerships head at a sponsor bank
A sponsor bank that treats fintech compliance as a pre-launch check will become the bottleneck. Require partners to map applicable rules to product requirements and audit evidence before adding accounts, cards or transfers.
Nubank completed deployment of Cardamon’s AI compliance platform across every jurisdiction where it operates. Cardamon reads applicable rules, identifies obligations for a given institution and maps them to product requirements; it says this can be completed up to 20 times faster than conventional systems or manual reviews. Nubank has more than 140 million customers and operates across Brazil, Mexico and Colombia. In the United States, it offers products through a partner-bank structure, with deposits through FDIC-insured Lead Bank, while its Federal Reserve and FDIC applications remain pending.
→ Action
Fintech partnerships: Require product requirements and audit evidence that map applicable rules before approving new partner products.
Full story →Read article → from Crowdfundinsider
Coinbase fine-tunes Qwen for fraud prevention
For the fraud head at a regional bank
A smaller model trained on your own fraud data may beat an expensive general model, but Coinbase’s internal benchmark is not proof for your payment flows. Require independent tests of recall, dollar-weighted recall, drift and block decisions before changing screening decisions.
Coinbase said it fine-tuned the open-weight Qwen3.5-9B model on proprietary internal data for fraud prevention in Onramp. Onramp supports crypto purchases in partner apps, including guest checkout, where users have little history; an existing machine-learning model and rules engine screen payments before an AI agent reviews selected transactions. The agent can block a payment that passed the first layer, but cannot reverse an earlier block. Coinbase said its internal benchmark showed Qwen beat Opus 4.5 by 7.5 points in precision, 12.0 points in recall and 35.4 points in dollar-weighted recall. Training used one NVIDIA RTX PRO 6000 96GB GPU for under $100, while median live-traffic latency was 55% lower than Opus 4.5.
→ Action
Model risk: Set independent tests for fraud-model recall, dollar-weighted recall, drift thresholds and override rates before approving a fine-tuned model.
Full story →Read article → from Coingabbar
General AI
Large language models & AI infrastructure
Salesforce agrees to acquire Loops
For the CIO at a midsize bank
If your bank uses Salesforce or Tableau, dashboards will no longer be a safe boundary. You need rules for data lineage, agent explanations and who can act on a recommendation before analytics moves into Slack and email.
Salesforce signed a definitive agreement to acquire Loops, an AI-powered analytics platform, and plans to bring its technology and specialized talent into Tableau after closing. Loops continuously monitors business performance, uses machine learning models to explain changes in key metrics and surfaces opportunities to improve. Its headless capabilities can run from Slack and email, monitor key performance indicators, identify trends and anomalies, and surface what is driving change. Salesforce says the acquisition will help customers identify business drivers, prioritize actions and move from insight to action in the flow of work. The transaction is expected to close in early November of Salesforce’s fiscal year 2027, subject to customary closing conditions.
→ Action
Data governance: Define lineage, explanation and approval requirements before connecting Tableau analytics agents to Slack or email.
Full story →Read article → from Salesforce
Get this in your inbox every morning
Free · No spam · Unsubscribe anytime