German bank lets ChatGPT and Claude trade for clients

Scalable Capital's move to integrate AI assistants into its trading platform is a significant step in bringing AI directly into financial workflows. The company, founded in 2014, has grown into one of Europe's largest digital investment platforms, with more than €60 billion in client assets and over 1 million customers. By using the open Model Context Protocol (MCP), Scalable avoids a formal partnership with AI providers, instead offering the connection as a standalone service that leverages existing AI capabilities. This approach allows the bank to maintain control over security and compliance, as users must approve every trade, and AI agents cannot access payment or withdrawal functions.

The Elm Wealth study from June provides a cautionary note. While AI models like Claude and ChatGPT showed skill in predicting market direction, they struggled with position sizing, taking on excessive risk relative to trade context. The study's authors highlighted that average position sizes of 7x to 12x could lead to catastrophic capital loss. Scalable's Chief Product Officer Alexander Siepp acknowledged that the integration may not be adopted at the same speed across all client segments, but he sees potential for a large group of interested clients. The company's infrastructure, which includes regulated banking services and strong customer authentication, is designed to allow investors to start their financial journey within an AI assistant and complete it securely through Scalable.