Goldman Sachs Bans Non-Engineers From Advanced AI Coding Tools

Goldman Sachs has banned traders, bankers, and other non-technical staff from using advanced AI coding tools, citing risk and productivity concerns. The restriction, reported by Bloomberg on November 2, limits access to tools such as Anthropic's Claude Code and Microsoft's GitHub to engineers and so-called "strats" — the investment banking term for workers who specialize in quantitative engineering and analytics.

The move has frustrated traders. Bloomberg reports they used the tools to monitor data, build financial models, and produce graphics. Those workflows now run through a narrower set of hands.

What does the restriction contradict?

The ban appears to undercut Goldman's own public posture on AI. In July, Goldman chief information officer Marco Argenti praised nontechnical staff for using AI to build software prototypes, then passing them to engineers for refinement. The bank's practice now points in the opposite direction: prototype freely, but only engineers keep the tools.

A Goldman spokesperson told Bloomberg the firm allows staff to experiment with certain AI tools "to determine the best use cases across different parts of the firm" and to assess "a variety of factors including productivity and risk management." Once the experimental period ends, "we put [the tools] in the hands of those who are best suited to leverage them for the long term," the spokesperson said.

Fast Company reached out to Goldman for additional comment but had not received a response by the time of publication.

How is the rest of Wall Street handling AI risk?

Goldman's decision fits a wider pattern. Wall Street institutions have been wrestling with how to balance speed against safety in AI adoption, and senior executives are split on where to draw the line.

JPMorgan Chase Chief Information Officer Lori Beer said last month that AI is evolving so quickly that banks struggle to assess the risks in real time. "We want to move at speed, but sometimes we must pull back to get controls in place," Beer said. "The technology is running ahead of capabilities to manage it."

Bank of America CEO Brian Moynihan has urged caution about letting AI agents operate without human supervision. Speaking in September, he said investment bankers need to carefully check the accuracy of AI-generated pitchbooks and spreadsheets. "The risk [is] letting it start getting answers without humans checking to make sure the answer was right," Moynihan said.

What is Goldman still building internally?

The ban does not signal retreat from AI. Goldman's internal initiatives include:

• A proprietary chatbot, available to all employees
• A "copilot" coding tool for developers
• A pilot of an autonomous coding agent, running as of September 2025
• A copilot tool for investment bankers, designed to help them compile research

That portfolio shows the bank's strategy in practice: buy control over the tools, then decide who gets to touch them. Goldman is not slowing its AI adoption — it is concentrating access in the hands of people it can audit, while senior bankers absorb the message that AI output requires verification.

The open question is whether traders and bankers get the tools back. Goldman's own framing suggests the door is ajar: the current arrangement, in the spokesperson's words, is an experimental period, not a permanent settlement.