OpenAI launched ChatGPT for Financial Services, a specialised workspace shaped through design partnerships with Morgan Stanley and Evercore for financial research, modelling and client materials. Its combination of financial data, firm templates and enterprise controls brings OpenAI directly into work previously divided across data terminals, productivity software and internal bank systems.

Elsewhere, AI moved further into payments, account access and bank operations. Ant International expanded its agentic payment protocol across wallets and acquirers, Meta connected Muse to financial accounts through Plaid and checkout through Stripe, five major banks invested in Rogo and Scotiabank quantified the time saved through its enterprise AI deployment.

Read more on the week’s key developments:

1. OpenAI combines financial data, modelling and client workflows in specialised ChatGPT

OpenAI launched ChatGPT for Financial Services, combining GPT-6 Astra with built-in data from Daloopa, PitchBook, LSEG News and Crunchbase. Financial institutions can connect existing subscriptions from providers including S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva and Moody’s. Administrators can publish firm-approved Excel, Word and PowerPoint templates, while compliance teams can export workspace logs into audit and investigation processes. The product initially targets investment banking and equity research and was shaped through design partnerships with Morgan Stanley and Evercore.

OpenAI is entering a market where Google recently introduced Gemini Enterprise for Financial Services with Deutsche Bank and CME Group as design partners. Google emphasises governed connections to institutional data and agents equipped with more than 50 financial skills, while OpenAI combines hosted datasets, financial reasoning and document production within ChatGPT. Competition is increasingly centred on which platform can connect licensed data, internal knowledge and firm templates while keeping outputs traceable and access controls intact.

2. Ant International expands AMP across 10 wallets and seven acquirers

Ant International began the global rollout of its Agentic Mobile Protocol across the Alipay+ ecosystem, which connects more than 50 mobile payment partners, over 10 national QR networks, 150 million merchants and two billion consumer accounts. The first phase includes Alipay, AlipayHK, DANA, GCash, KakaoPay, MPay, TNG eWallet, TrueMoney, Toss and Starryblu. Adyen, Allinpay, Checkout.com, Fiserv, Global Payments, Nuvei and Worldline are participating on the acquiring side. Ant also made AMP source code, development kits and technical documentation available on GitHub.

Mastercard and Visa are working with Ant International on a Know-Your-Agent interoperability framework through BuildFin.ai, an industry platform convened by the Monetary Authority of Singapore. AMP separates authorisation of a task from access to the underlying account and proposes agent identities, certified capabilities and trust ratings to control levels of autonomy. With wallets and acquirers joining the first phase, the focus moves to how those permissions and liability protections function when an agent crosses payment networks, devices and merchant environments.

3. Meta’s Muse connects financial accounts through Plaid and checkout through Stripe

Meta launched Muse with Plaid-powered financial-account connections, allowing US users to share balances, transaction histories, investment holdings and mortgage information with the personal AI agent. Plaid connects to more than 12,000 financial institutions and financial applications in the US, with users choosing which accounts to share and retaining the ability to revoke access. Muse can also complete approved purchases through Stripe’s Link infrastructure, including at participating merchants and through single-use virtual cards where required.

Muse gives a general-purpose AI agent visibility across a consumer’s financial position before it recommends or prepares an action. Banks can retain the account and regulated balance-sheet relationship while more of the customer’s budgeting, product discovery and transaction journey occurs through an external interface. This raises the commercial importance of permission controls, data quality and how prominently the underlying institution remains visible within agent-led interactions.

4. Barclays, BNP Paribas, Citi, MUFG and Société Générale invest in Rogo

Barclays, BNP Paribas through Opera Tech Ventures, Citi Ventures, MUFG Innovation Partners and Société Générale invested in Rogo, joining earlier strategic investors including JP Morgan Growth Equity Partners and Truist Ventures. New York-based Rogo provides a model-agnostic AI platform for financial research, data analysis and document workflows. The company reports deployments covering more than 50,000 professionals across over 350 firms and plans to use the investment to expand its EMEA and APAC operations.

The latest investors bring the number of global banks on Rogo’s shareholder register to nine, with nearly $20 trillion in combined assets. Their participation gives Rogo institutional backing as larger providers such as OpenAI and Google develop finance-specific platforms. Specialist firms retain an opening where banks require technology built around financial data, firm knowledge and established analyst workflows, although they will increasingly need to integrate with the broader AI environments banks adopt.

5. Mistral raises EUR 3 billion as HSBC and BNP Paribas deploy its models

Paris-based Mistral AI raised EUR 3.0 billion ($3.5 billion) in a Series D round led by Samsung Electronics, with a post-money valuation exceeding EUR 21.0 billion ($24.5 billion). The three-year-old company plans to expand model research, computing capacity, infrastructure and international operations. It operates in 20 countries and supports more than 125 enterprises, including HSBC, which has a multi-year partnership with Mistral. BNP Paribas also uses its models across markets, sales and customer support.

Mistral is building its commercial case around open-weight models, private deployment and greater control over where data and computing reside. These features are relevant to banks handling confidential client information across jurisdictions with differing data and regulatory requirements. The funding gives Mistral more capacity to compete with US providers while offering institutions another model and infrastructure option within a diversified AI architecture.

6. BIS paper says frontier AI is compressing banks’ cyber-response windows

A Financial Stability Institute paper published by the Bank for International Settlements found that frontier models can autonomously identify vulnerabilities, develop exploits and conduct complex multi-step cyber operations. The technology reduces the expertise and time required for sophisticated attacks and can compress the interval between discovering and exploiting a weakness. The same capabilities can help defenders identify vulnerabilities and respond to incidents more quickly.

The immediate technology risk for banks is a widening mismatch between automated vulnerability discovery and the time required to validate findings, test patches and implement changes across critical systems. Existing cyber-resilience practices remain applicable but need to operate at a faster tempo. Banks will require clearer vulnerability prioritisation, shorter patching cycles and stronger coordination with cloud, software and infrastructure providers while preserving controls around changes to important services.

7. Krungsri brings AI-enabled operations into five-year core banking programme

Krungsri, formally Bank of Ayudhya, renewed its technology partnership with Kyndryl for five years. Kyndryl will support the Thai bank’s core environment and infrastructure covering more than 4,800 ATMs through its Bridge platform, automation, Mainframe-as-a-Service and application-modernisation services. Agentic AI will be used for application and infrastructure analysis, operational automation and selected modernisation work under human oversight.

Krungsri is Thailand’s fifth-largest financial group and one of its six domestic systemically important banks, with MUFG as a strategic parent. Applying AI to infrastructure monitoring, batch processing and application analysis puts the technology inside the systems supporting everyday banking services. The five-year programme also reflects the pace required to modernise a large bank’s core environment while keeping legacy services available and resilient.

8. MAS details bank controls against AI-enabled impersonation scams

The Monetary Authority of Singapore outlined safeguards introduced with banks following scams involving AI-generated impersonation, including a case with losses of around SGD 5 million ($3.9 million). Banks now detect rapid account draining and can block or hold subsequent transfers. They have also introduced delays of at least 12 hours for higher-risk actions such as adding new payees or increasing transfer limits, while PayNow nicknames have been removed to reduce impersonation.

MAS is combining transaction controls with measures covering the platforms through which scams begin, including new codes for messaging and conferencing services and a recognisable prefix for government calls. Visa’s revised Singapore security roadmap adds industry measures including tokenisation, passkeys, biometric authentication and more than 150 AI and machine-learning models used globally for fraud detection and authentication. The approach concentrates additional friction around higher-risk actions while maintaining ordinary digital-payment access.

9. Goldman Sachs opens AI and cloud engineering office in Bellevue

Goldman Sachs opened an engineering office in Bellevue, Washington that can accommodate more than 125 employees working on artificial intelligence and cloud transformation. It is the firm’s first dedicated engineering space in the region and complements an existing Seattle office serving its Banking and Wealth Management businesses. Goldman employs more than 12,000 engineers globally, representing approximately one-quarter of its workforce.

At 125 seats, Bellevue is a targeted addition to Goldman’s global engineering base. Its location gives the firm access to experienced cloud and AI engineers in one of the largest US technology centres, supporting internal development alongside the external platforms and models it uses. The expansion continues Goldman’s investment in proprietary technical capacity as AI becomes more closely integrated with software development, data and client-facing businesses.

10. Scotiabank’s AI applications save 24,000 workdays in four-and-a-half months

Scotiabank President and CEO Scott Thomson said the bank’s AI applications saved approximately 24,000 workdays over four-and-a-half months. The disclosure adds a measurable productivity outcome to Scotia Intelligence, the bank’s enterprise approach to applying AI across client service, operations, software development and decision-making.

Scotiabank said in June that more than 71,000 employees had access to assistive AI tools, including 5,500 engineers using AI to support coding. Employee use of AI to respond to client questions had increased 30% quarter on quarter. Its latest results showed that higher technology costs contributed to a 9% increase in adjusted expenses, while the adjusted productivity ratio improved from 53.7% to 52.5%. The workday figure provides a more direct operational measure alongside these bank-wide results.