BankingNewsAI Daily Brief  · 

Google launched Gemini Enterprise for financial services, with Deutsche Bank as an early adopter.

🏦 3 Banking AI🤖 3 General AI

Banking AI

Financial institutions & fintech technology

3 stories
cloud.google.com

Google launched a financial-services version of Gemini Enterprise, with Deutsche Bank as an early adopter

Google Cloud introduced Gemini Enterprise for Financial Services, a purpose-built agentic platform for financial workflows. Deutsche Bank has signed on as an early adopter, and Dun & Bradstreet is supplying verified commercial-data context for onboarding, lending and underwriting use cases.

Action

Pressure-test Google’s financial-services offering against existing bank AI architecture, particularly its controls for grounded research, data entitlements and workflow-level auditability.

Read article →
housingwire.com

Colorado’s proposed automated-decision rules would add mortgage AI disclosures, documentation and human review

Colorado’s proposed ADMT rules would apply new documentation, disclosure and human-review obligations where technology materially influences lending decisions. Mortgage industry groups are challenging definitions and scope, but the proposal makes automated underwriting governance a near-term state compliance issue.

Action

Map every model and vendor tool that affects mortgage eligibility, pricing or adverse-action workflows against the proposal now; preserve decision records, human-override evidence and consumer-notice capability.

Read article →
housingwire.com

UWM put mortgage-broker discovery and affordability tools inside ChatGPT

United Wholesale Mortgage launched a Mortgage Matchup plugin for ChatGPT that helps consumers locate independent brokers, run affordability calculations and access mortgage education. The move shifts an early stage of borrower acquisition into an AI-agent interface rather than a lender-controlled site or app.

Action

Treat AI assistants as an emerging mortgage-distribution channel: define what approved product, pricing and educational content an external agent can access and how referrals are measured and supervised.

Read article →

General AI

Large language models & AI infrastructure

3 stories
news.smol.ai

DeepSeek added multimodal agents at flash-tier pricing, increasing pressure on US bank AI vendors

DeepSeek released V4-Flash-Vision-Exp with mixed text-and-image API support, a reusable Files API, and image-token billing at Flash pricing. DeepSeek says its multimodal-agent performance approaches Opus-4.8, making high-volume document, image, and workflow automation materially cheaper for enterprises that can use the service or comparable vendor offerings.

Action

Benchmark this model class against your current document-intelligence and agent vendors for image-heavy workflows such as KYC document review, claims, and back-office exception handling.

Read article →
openai.com

OpenAI published first results for its custom Jalapeño inference chip

OpenAI says Jalapeño delivers higher throughput and lower latency for modern-model inference while using power more efficiently. The announcement is a meaningful signal that leading model providers are moving deeper into custom inference hardware, where serving cost and response time increasingly determine enterprise AI economics.

Action

Recalculate multi-year AI-unit-cost assumptions around faster, cheaper inference rather than model-license pricing alone; require strategic vendors to disclose how their infrastructure roadmap affects latency, capacity and price.

Read article →
research.ibm.com

IBM released Granite 4.2 reasoning models under Apache 2.0

IBM released Granite 4.2 in 3B, 8B and 30B sizes, adding native reasoning for enterprise agents and publishing the weights under Apache 2.0. The release gives enterprises an openly licensed option for workloads where self-hosting, controlled tuning and data residency matter more than frontier-model breadth.

Action

Benchmark Granite 4.2 for constrained internal-agent use cases—especially policy retrieval, operations and developer tooling—where open weights can reduce vendor concentration and improve deployment control.

Read article →

Get this in your inbox every morning

Free · No spam · Unsubscribe anytime

Subscribe free →