Goldman Sachs Cuts Off Traders and Bankers from AI Coding Tools

Goldman Sachs has abruptly limited access to advanced artificial intelligence coding tools, revoking permissions for the majority of its non-technical staff. Under internal guidelines issued in October, the investment bank now restricts platforms such as Anthropic's Claude Code and Microsoft's GitHub exclusively to software engineers.

The policy shift removes access for traders, bankers, and asset managers who had spent recent months deploying the software to automate tasks, monitor trading flows, and generate content. The sudden rollback has reportedly sparked frustration across several business units within the firm.

Reversing the "Vibe Coding" Strategy

The clampdown marks a sharp contrast to the bank's previous messaging regarding generative AI adoption. In July, Goldman Sachs Chief Information Officer Marco Argenti publicly praised the concept of "vibe coding" among non-technical personnel.

Under that vision, traders and business analysts were encouraged to use AI assistants to rapidly build initial prototypes of their ideas, which would then be handed over to dedicated engineering teams to refine and scale for production environments. That approach now appears to have been shelved.

Shifting from Experimentation to Control

Addressing the restriction, Goldman Sachs spokesperson Nick Carcaterra stated that the policy change stemmed from a "variety of factors including productivity and risk management."

According to Carcaterra, the institution has formally moved past its initial exploration phase. The firm is now narrowing tool availability strictly to specialists who are "best suited to leverage them for the long term."

While Wall Street institutions have raced to integrate generative AI to gain an edge, Goldman Sachs' latest move underscores the rising enterprise concerns over compliance, data safety, code reliability, and oversight when deploying powerful AI tools broadly across non-technical divisions.