Goldman Sachs Limits AI Coding Tools to Its Engineers
Goldman Sachs is limiting access to AI coding tools such as Claude Code and GitHub Copilot to engineers, citing productivity and risk management as the reasons for the move.
Goldman Sachs is withdrawing access to AI-powered programming tools such as Claude Code and GitHub Copilot from traders and bankers. In future, only the bank's engineers will be able to use these tools. The decision marks a clear reversal, as the tools had previously been deployed more widely within the firm.
As recently as July, CIO Marco Argenti publicly praised so-called vibe coding, in which non-technical employees create software with the help of AI. Argenti stressed at the time that this approach opened up new possibilities. That the bank is now scaling back the same approach shows how quickly internal assessments have shifted.
Goldman Sachs cites "productivity and risk management" as the reasons. The bank thus points to two competing tensions: on the one hand, AI tools are meant to speed up work; on the other, concerns are growing about errors, security vulnerabilities and uncontrolled code when employees without solid programming skills generate software. By restricting access to engineers, the institution aims to manage these risks more effectively.
The move illustrates how financial firms are currently balancing the use of generative AI. While many companies want to roll out the technology as broadly as possible, banks whose systems affect trading, settlement and customer data are opting for tighter controls. Whether other institutions will follow suit remains open; Goldman Sachs' decision remains the central finding so far.
Source: the-decoder.com
9
