T-Investments Opens AI Trading Tool to All Clients

The tool works through the Model Context Protocol, or MCP, and lets users connect agents such as Claude, DeepSeek and ChatGPT to a brokerage account. Clients can issue plain-text instructions — for example, to rebalance a portfolio or buy a stock at a target price — and the agent translates them into trading orders. Access can be limited to viewing balances and transaction history, or expanded to execution and transfers between a user’s own accounts.

T-Investments said responsibility for decisions remains with the investor, and that any output from the agent — including stock selection or news sentiment summaries — will be treated as a technical digest rather than personalised investment advice. The company also said the system cannot act outside an active chat session and will require explicit confirmation before any trade goes through.

That warning matters because the commercial appeal of AI-led investing is tied to convenience, but the legal and operational risks are tied to miscommunication, overreach and cybersecurity. By framing the tool as a technical assistant rather than an adviser, the broker is trying to reduce liability while still giving clients a more autonomous trading workflow. The balance is important: the more seamless the user experience, the greater the pressure on controls, disclosures and client understanding.

For T-Bank, the move also fits a wider industry race in which banks and brokers are racing to show they can operationalise AI without compromising trust. Recent filings from large US brokers and wealth managers have highlighted cybersecurity and technology risk as major concerns, and the same logic applies here: once an AI agent can access live accounts, even under guardrails, the platform becomes more dependent on authentication, permissions management and execution integrity.

The near-term commercial upside is clear. Mass availability can lift usage among self-directed investors, encourage more frequent trades and help the firm differentiate its web platform from rivals still treating AI as a feature rather than a product layer. The bear case is that the service may attract regulatory scrutiny if users treat it as advice or if clients misjudge the scope of an agent’s authority. That makes the rollout less about novelty than about whether the brokerage can scale automation while proving the human remains in control.