Goldman Sachs narrowed use of Anthropic’s Claude Code and Microsoft’s GitHub coding tools under an October guidance update, Bloomberg reported, with engineers retaining access and “strats” exempt from the restriction.
Bloomberg said bankers and traders had used the tools for several months before the October change. Its account of the guidance said engineers could use them, while “strats” were exempt; those staff combine work in trading, quantitative research and software engineering. The outlet did not state how many employees were affected or specify when the new guidance took effect.
Some traders used the software to track client trading flows, model market movements and positions, and prepare written and visual material, Bloomberg reported. The outlet cited a person familiar with the matter who said some traders were frustrated by the change. Bloomberg gave no measured productivity gains, cost savings or risk incidents linked to the tools or review.
In comments to Bloomberg, Goldman spokesperson Nick Carcaterra said the firm weighs how tools perform and what risks they pose after a trial period before setting ongoing access. He did not detail the findings on Claude Code or GitHub.
The reported limits coexist with Goldman’s broader AI effort. In a Nov. 4 interview published by the bank, President and Chief Operating Officer John Waldron described employee AI literacy and adoption, along with software development, as parts of Goldman’s AI strategy. He said coding tools had improved efficiency and productivity for the firm’s coders, but did not identify the tools or connect those remarks to Claude Code and GitHub. Waldron did not discuss the access change.
Goldman’s 2025 annual report warned that new AI technologies may increase the frequency and severity of cybersecurity attacks. It presents this as a general risk disclosure and does not link it to the reported access decision.
Bloomberg also reported that Goldman had restricted Hong Kong staff from using Anthropic’s Claude earlier in 2026.
