Goldman Sachs Is Pulling Claude Code from Its Bankers and Traders (www.thestateofai.com)

Goldman Sachs has restricted the use of Anthropic's Claude Code and Microsoft's GitHub solely to its engineering staff, limiting access for traders and bankers who had previously utilized these AI coding tools for tasks like monitoring client trading flows and constructing market models. This decision, rooted in concerns about productivity and risk management, follows a review that highlighted the potential dangers of traders using unvetted code that could handle sensitive data without sufficient oversight. While engineers maintain access to these tools, many traders express frustration over the sudden restriction, especially after Goldman’s CIO praised their contributions to AI tool development just a few months prior. This policy change is significant for the AI/ML community as it underscores the ongoing tension between innovation and regulatory compliance in high-stakes environments like finance. The restriction serves as a reminder of the inherent risks of using AI-driven coding agents that operate without oversight, particularly in roles that handle sensitive information. By requiring that only qualified engineers use these tools, Goldman aims to enhance control over the integrity and security of its coding processes while also positioning itself strategically ahead of Anthropic's upcoming IPO, where it serves as a lead underwriter.