Goldman Sachs Just Pulled Claude Code From Its Bankers and Traders — Here Is the AI Coding Governance Audit Service That Follows From It.

Goldman Sachs spent months letting traders and bankers use Claude Code and GitHub to prototype tools.

In October 2026, it took that access away from almost all of them.

What actually happened

Internal guidance updated in October 2026 restricted Claude Code and GitHub access to Goldman's roughly 12,000 engineers, pulling the tools from most traders and bankers who had been using them to monitor client trading flows, simulate market moves and positions, and draft text and visuals.

A Goldman spokesperson framed the change plainly: "After a period of experimentation where we assess a variety of factors including productivity and risk management, we put them in the hands of those who are best suited to leverage them for the long term."

One group kept access: "strats," Goldman's hybrid staff who combine software engineering, quantitative research, and trading in a single role. Everyone outside engineering and that narrow hybrid category lost direct access.

Why this is a business, not just a headline

The traders and bankers who lost access weren't misusing the tools maliciously — they were doing exactly what every non-engineering team at every company is doing right now: building quick prototypes with an AI coding agent, then handing them off (or not) for review. Goldman's own stated reasoning was a mix of productivity and risk management, which is a nice way of saying: unreviewed code was running in places that mattered, and nobody had mapped who was writing it.

Almost no company outside Goldman has formalized that mapping. Most just let access spread with whoever requested a license.

The pitch that sells itself

Open with Goldman's own move, not a generic security warning: "Goldman Sachs just pulled Claude Code from most of its traders and bankers after a risk review — keeping it only for engineers and one narrow hybrid role. We'll run the same inventory on your team, free for one department, before you decide on the full audit."

That's concrete, matches a decision a name-brand bank already made, and gives the client a way to verify the finding themselves.

Bottom line

This works as a playbook because the policy gap isn't hypothetical — Goldman just confirmed it exists even inside a bank with a dedicated risk function. Mapping who has unreviewed access to AI coding agents, and writing the policy that narrows it, is a service almost every regulated or risk-conscious company needs and none of them have built yet.

Source: https://dealroom.co/news/169162-goldman-restricts-claude-code-github-for-traders-and-bankers/