The Pulse
Goldman Limits AI Coding Tools to Engineers
Goldman Sachs is restricting traders and bankers from using Anthropic’s Claude Code and Microsoft’s GitHub coding tools, according to internal guidance updated in October. The change follows months of wider access and leaves an exception fo
AI.info Team · Written with AI · fact-checked ·
Goldman Sachs is narrowing access to AI coding tools that traders and bankers had used to build work applications, limiting Claude Code and Microsoft’s GitHub to engineers, according to internal guidance updated in October. The restriction reverses months of wider access at the bank, where some nontechnical staff used the tools to monitor client trading flows, model market scenarios and create written and graphic material.
From wider access to an engineering boundary
The change applies to Anthropic’s Claude Code and Microsoft’s GitHub, which Goldman had allowed bankers to use for work. The revised guidance limits the tools to engineers, according to Bloomberg’s report. Employees known as “strats,” whose jobs combine trading, quantitative research and software engineering, are exempt from the restriction, a person familiar with the decision told Bloomberg.
The move has frustrated some traders who had used the assistants for practical tasks, including monitoring client flows and modeling market changes and positions. Those uses extend beyond writing software: employees could use coding assistants to assemble tools for recurring analysis and produce other work materials. The report does not say that Goldman has barred non-engineers from every form of AI.
Goldman’s experiment meets risk controls
Goldman spokesperson Nick Carcaterra described the restrictions as the result of a trial period in which the firm assessed productivity and risk management. “We make certain AI tools available to our people to determine the best use cases across different parts of the firm,” he said. The bank’s position, as described by Carcaterra, is that access should sit with employees best suited to use the tools over the long term.
The change follows a period when Goldman’s technology chief, Marco Argenti, publicly praised nontechnical employees’ use of AI to make early versions of tools. In July, Argenti said traders and bankers had built solutions with AI and passed them to the firm’s 12,000 engineers for improvement. The new guidance draws a clearer line between making an initial tool and having broad access to coding assistants.
A narrower policy, not a firmwide AI ban
Goldman has also developed an internal AI assistant that combines models from OpenAI, Anthropic and Google’s parent company, Alphabet. The coding-tool restriction therefore does not amount to a withdrawal from AI use across the bank; it changes which employees can use particular third-party tools. Goldman had previously restricted Anthropic’s Claude for staff in Hong Kong, a separate location-specific decision.
Other banks are weighing how to match AI tools to the work employees need to do. JPMorgan Chase said in July that it wanted staff to consider which model suited a task, rather than use a more expensive option for something as straightforward as summarizing an analyst report. At Goldman, the immediate boundary is clearer: engineers retain access to the coding tools, while traders and bankers generally do not, with strats carved out.
The strats exception marks the line
That exception points to the problem behind any job-based restriction: some financial roles already involve building and maintaining software. Goldman’s policy separates employees by function, but the strats carveout acknowledges that the boundary is not clean. The available account does not specify how the bank will assess individual access requests or whether the restrictions will change after further review.