Goldman Sachs has restricted access to advanced AI coding tools, including Anthropic's Claude Code and Microsoft's GitHub, to its engineers. The bank updated internal guidance in October to limit these tools to approximately 12,000 engineers, barring traders and bankers who had previously used them for tasks like tracking client flows and building market models. This move follows a review of productivity and risk concerns, including information-barrier requirements and unreviewed code on trading desks.
The change signifies a shift from Goldman Sachs' earlier embrace of AI experimentation. In July, Chief Information Officer Marco Argenti had praised "vibe coding," where non-technical employees built AI prototypes. However, the October guidance restricts these powerful AI agents to employees with engineering expertise, a move that has reportedly caused frustration among some traders. Specialists known as "strats," who work across trading, quantitative research, and engineering, retain access to these tools. This development highlights the ongoing challenge financial institutions face in balancing AI innovation with operational and regulatory risks.
Goldman Sachs has previously implemented AI tool restrictions, notably limiting Claude's use for its Hong Kong-based bankers in April. That earlier restriction was location-based, whereas the current policy is role-based and applies firm-wide. The bank stated that it makes certain AI tools available to staff to determine the best use cases and assess productivity and risk factors. After this experimental period, the tools are placed "in the hands of those who are best suited to leverage them for the long term." This approach suggests a move toward more controlled deployment of AI technologies within the firm.
The decision comes weeks before Goldman Sachs is set to lead the initial public offering for Anthropic, the developer of Claude Code. This timing raises questions about potential conflicts of interest, though the bank has maintained its close working relationship with Anthropic. The bank's internal AI initiatives also include a proprietary chatbot accessible to all employees and a "copilot" coding tool for developers. Goldman Sachs continues to invest in AI, but this recent policy change indicates a more cautious and segmented approach to its adoption.
This move by Goldman Sachs reflects a broader trend on Wall Street, where financial institutions are grappling with the rapid evolution of AI. While the potential benefits are substantial, concerns about sensitive data, security risks, and regulatory compliance are prompting a more measured deployment strategy. The bank's spokesperson, Nick Carcaterra, emphasized that the firm offers AI tools for exploration to identify optimal use cases, after which they are given to employees best positioned for long-term utilization. This strategy aims to balance innovation with the need for robust controls in a highly regulated industry.