Loading page…
The restriction follows July praise for staff-built prototypes. Some traders are frustrated, while Goldman cites productivity and risk management in deciding who gets access.
Listen to this story
Goldman Sachs is narrowing access to AI coding tools after months of experimentation: October guidance, reported by Bloomberg on November 2, 2026, says bankers and traders can no longer use tools such as Claude Code and GitHub, while engineers retain access. The shift limits nontechnical staff’s ability to turn ideas into prototypes before engineering handoff, though employees also used the tools for other work. Goldman says it is selecting long-term users after weighing productivity and risk; it disclosed no specific failure or comparative results.
In July, CIO Marco Argenti praised nontechnical “vibe coding,” with prototypes intended to pass to one of Goldman’s 12,000 engineers for production work.
Some traders are frustrated; reported use cases included monitoring client trading flows and creating written and graphic materials.
Goldman spokesperson Nick Carcaterra said the access period assessed productivity and risk management, but provided no metrics or specific trigger for the restriction.
Goldman Sachs bankers and traders are losing access to AI coding tools they had used for months, while engineers can keep using them. Bloomberg reported the restriction on November 2, 2026, citing internal guidance issued in October. The change covers tools including Anthropic’s Claude Code and Microsoft’s GitHub, according to PYMNTS’ account of the reporting.
As recently as July, Goldman Chief Information Officer Marco Argenti had praised “vibe coding” by nontechnical employees: using AI to build software without doing all the coding themselves. A trader or asset-management employee could turn an idea into a quick prototype, then hand it to an engineer to prepare for production, The Decoder’s account explains. Argenti described those creations going to one of the firm’s 12,000 engineers for enhancement, giving business staff a role in developing solutions before technical specialists took over.
Some traders are frustrated by the restriction, a person familiar with the matter told Bloomberg. Their uses extended beyond software experiments: they had employed AI to monitor client trading flows and create written and graphic materials. Those examples describe how employees used the tools, rather than measured gains in productivity.
Goldman spokesperson Nick Carcaterra described the initial access as a way to determine the best uses for certain AI tools across different parts of the firm. He framed the subsequent choice of users as part of that evaluation:
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.
Nick Carcaterra, Goldman Sachs spokesperson, speaking to Bloomberg, as quoted by PYMNTS
Carcaterra’s explanation did not identify a specific failure that prompted the restriction or provide figures comparing productivity among engineers and other employees. The bank’s stated rationale leaves those results unspecified.
Loading discussion...
Join the conversation
Explain where you would draw the line between experimentation and restricted access.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.