AI in action · Tools
Goldman Takes AI Coding Tools Back From Its Traders
Goldman Sachs has limited AI coding tools such as Claude Code to its engineers, cutting off bankers and traders, Bloomberg reported on November 2. Risk management was one reason.
For most of this year, the fastest-growing use of AI on Wall Street trading floors was not a trading model but a coding assistant. At Goldman Sachs that experiment has now been narrowed. Under internal guidance issued in October and reported by Bloomberg on Monday, November 2, advanced AI coding tools, including Anthropic's Claude Code and Microsoft's GitHub tools, are restricted to the firm's engineers. Bankers and traders who had used them for months no longer have access.
What the traders were doing with them explains both the appeal and the concern. According to Bloomberg's source, they used the tools to monitor client trading flows and to produce written and graphic materials: small programs written by people who are not software engineers, running close to client data and trading decisions. In July, Goldman's chief information officer, Marco Argenti, had praised exactly this kind of "vibe coding" by non-technical staff, describing a model in which a trader or asset-management employee builds a quick prototype and hands it to one of the firm's roughly 12,000 engineers to make production-ready.
The bank's explanation was brief. Bloomberg cited "a variety of factors including productivity and risk management". Spokesperson Nick Carcaterra said the firm makes certain AI tools available "to determine the best use cases across different parts of the firm", and, according to The Decoder, that the experimentation phase is over and the tools will be limited to people "who are best suited to leverage them for the long term". A person familiar with the matter told Bloomberg that the change frustrated some traders.
The risk-management side is easy to see from a supervisor's chair. Code written by a trader with an AI assistant can be quick and useful, but it may not go through the testing, change control and record-keeping that a bank's own systems do, and regulators from the FCA to the Hong Kong SFC have spent the autumn telling firms that they stay accountable for whatever their AI tools produce. Restricting who can generate code is one of the simplest controls available.
The move is not a retreat from AI as a whole: engineers keep the tools, and nothing in the reports suggests the bank is scaling back its wider AI programme. It is a sign, though, that the first wave of open access is giving way to rules about who may use which tool, and for what. Other banks that let front-office staff build their own software with AI will face the same choice.
Sources: PYMNTS — Goldman Sachs bars bankers from using Claude Code and GitHub (Nov 2, 2026); The Decoder — Goldman Sachs pulls AI coding tools from traders and bankers (Nov 2, 2026)