European investment platform opens trading to AI assistants, with mixed results in early tests
Scalable Capital, a Munich-based bank, has become the first in Europe to let customers use AI assistants like ChatGPT and Claude to manage portfolios and place trades. The company claims this democratizes access to financial intelligence, but a June study found AI performance varies, with Claude outperforming human traders in 76% of cases. The service is available to over a million clients across Europe.
Scalable Capital’s integration relies on the Model Context Protocol, a technical standard that lets external AI agents securely interact with its banking systems. The firm’s client base spans over a million users, with €60 billion in assets concentrated in German-speaking markets but expanding into Southern and Western Europe. Its July expansion to 1.8 million derivatives signals a broader push beyond simple stock trading into complex instruments.
The Elm Wealth study’s methodology is notable: AI models were given historical newspaper front pages but not outcomes, simulating real-world information gaps. While Claude and ChatGPT showed strong directional forecasting, their aggressive position sizing—often 7 to 12 times leverage—suggests a disconnect between theoretical risk formulas and practical execution. This mirrors a common challenge in algorithmic finance: knowledge does not automatically translate into prudent behavior.
This move could reshape retail investing by lowering barriers to sophisticated analysis, potentially benefiting novice traders who lack research skills. However, the risk of overconfident AI-driven trades may amplify losses, especially if users blindly follow prompts without understanding position sizing. Regulators might face pressure to clarify liability when AI advice leads to financial harm. The mixed test results suggest that while AI can democratize information, it does not yet reliably manage risk—meaning investors could gain access but still need human oversight to avoid catastrophic mistakes.