Goldman Sachs Takes AI Coding Tools Away From Traders and Bankers

Goldman Sachs has reportedly removed most employees’ access to advanced AI coding tools such as Anthropic’s Claude Code and Microsoft’s GitHub Copilot. According to Bloomberg, as relayed by The Decoder, only engineers can still use them under internal guidelines issued in October. Traders and bankers, who had used the tools for months, are said to be frustrated.

The detail that makes this notable is the timing. In July, the bank’s chief information officer, Marco Argenti, publicly praised “vibe coding” by non-technical staff. Vibe coding means describing what you want in plain language and letting an AI write the program. The idea was that a trader could build a quick prototype of, say, a tool that watches trading flows, then hand it to an engineer to make it production-ready. The report says that plan is no longer in place. A Goldman spokesperson cited “a variety of factors including productivity and risk management” and said the experimentation phase is over, with tools going to people “who are best suited to leverage them for the long term.”

What’s behind it: Banks run on rules about who may change what, and who is accountable when something breaks. A script written by a non-engineer with an AI helper can work well on the day and still be hard to review, test or maintain. That is not a flaw of AI as such; it is an old problem (so-called shadow IT, meaning software built outside the official process) arriving faster because building is now so easy. Goldman has not said which factors decided it, how many people are affected, or how “best suited” is judged, so we only have the outline.

It is also worth keeping the scale in view. This is one bank, reported by one outlet, and nothing suggests the bank is abandoning AI. Reading it as “AI coding failed” would be too strong. Reading it as “the easy part is writing code, the hard part is owning it” is closer to the evidence.

What this means for you: If you use AI to build small tools at work, check what your employer allows before you rely on them, especially if they touch customer data, money or compliance. Keep a note of what each tool does and who can maintain it, so it does not become a mystery when you change roles. If you are just curious, the takeaway is mild: AI makes it easy to create software, and organisations are now deciding who gets to do that and under which rules.

Sources

• The Decoder: Goldman Sachs pulls AI coding tools from traders and bankers