Goldman Sachs has restricted access to advanced AI coding assistants, limiting tools like Anthropic's Claude Code and Microsoft's GitHub exclusively to engineers. Non-technical staff, including traders and bankers who previously used the software to monitor trading flows and generate content, were cut off under internal guidelines issued in October. This rollback reverses the bank's earlier openness to broader experimentation among financial professionals.
The bank's Chief Information Officer Marco Argenti had previously highlighted the potential of non-technical staff using AI for rapid prototyping, where employees built quick models before handing them over to engineers. However, Goldman Sachs spokesperson Nick Carcaterra stated that the experimentation phase has ended. The bank cited a variety of factors including productivity and risk management for the U-turn, deciding to restrict the tools to personnel best suited to leverage them long-term.
This policy shift has reportedly caused frustration among traders and bankers who had spent months integrating the technology into their daily workflows. By narrowing deployment strictly to engineering teams, Goldman Sachs joins other financial institutions navigating the balance between AI innovation and operational risk.
Why it matters: The decision highlights how major financial institutions are stepping back from broad AI experimentation to impose stricter controls and risk management on software deployment.