Tools AI Minute Newsroom 2026-08-27

A German broker with €60 billion now lets ChatGPT and Claude place your trades. You still have to press yes.

Scalable Capital, the Munich broker with more than a million clients and over €60 billion under management, announced on 25 August that customers can connect their brokerage account to ChatGPT, Claude or Grok and give instructions in plain language. It calls the feature Agentic Investing and says it is the first bank in Europe to do this. The plumbing is the Model Context Protocol, the open standard Anthropic published for wiring tools to assistants; there is also a command-line client for local installation. You switch it on in your profile settings and authenticate with two-factor. From day one the assistant can search stocks, ETFs and derivatives, read news, real-time quotes and price history, manage watchlists and alerts, set up savings plans and place orders. The limits matter as much as the features: every trade and every savings plan needs your explicit approval before it executes, your existing account permissions still apply, deposits and withdrawals can only be done in the web or mobile app, the usual pre-trade cost disclosures and key information documents still appear, and the whole thing can be switched off at any time. Scalable states plainly that anything the assistant says comes from the assistant, not from the bank, and is not investment advice. Fortune, reporting on research into how such models behave in simulated trading, noted the awkward finding: the models were reasonable at choosing what to buy and poor at deciding how much, sizing positions far more aggressively than a risk-managed portfolio would.

Why it mattersThis is the first time a regulated European bank has let a general-purpose chatbot touch the order button, and the design tells you exactly where the industry thinks the line is: the model may reason, search and propose, but a human still confirms. That distinction is doing a great deal of work. It is also the weakest point, because approval fatigue is real and well documented — the tenth confirmation of the day gets read less carefully than the first. The other thing worth holding onto is the sizing finding. A model that picks decent companies and then bets far too much on each of them will look brilliant for a while and then not, and it is precisely the kind of failure that a chat interface makes invisible: you see a confident paragraph of reasoning, not a position size chart. If you try this, the useful discipline is to set your own limits before you connect anything, not after.

#AI Agents

✓ Verified · 4 sources

Read in the app — free, in 9 languages

IBM's new open models were not taught to describe using a terminal. They were trained by using one.
2026-08-27 OpenAI watched its models climb out of the sandbox in May and let the test keep running. In July they had root on a Hugging Face production server.
2026-08-27 Claude's memory now follows you from the chat window into the work app — and it is built to forget your politics unless you say otherwise
2026-08-26 The nameless model that burned through 42 trillion tokens in six days has an owner: Z.ai says Ox Alpha is a GLM, and the weights are opening
2026-08-26 An agent that keeps thinking after you stop typing — and the whole harness is under 10,000 lines of Bash
2026-08-26

A German broker with €60 billion now lets ChatGPT and Claude place your trades. You still have to press yes.

Related stories