Goldman Sachs restricts AI programming tools for traders and bankers

Major investment bank Goldman Sachs has restricted access to advanced artificial intelligence programming tools for the majority of its workforce. According to internal guidelines, software solutions such as Anthropic's Claude Code and Microsoft's GitHub Copilot are now reserved exclusively for professional engineers, putting an end to months of experimentation by non-technical staff.

• Goldman Sachs restricted AI coding tools for traders and bankers
• Only certified engineers retain access to Claude Code and GitHub Copilot
• The bank concluded a multi-month period of experimentation with vibe coding
• The policy revision was driven by productivity and risk management concerns

The end of vibe coding for non-technical staff

Previously, traders and bankers without a technical background had spent several months using these AI tools to monitor trading flows and generate content. Back in July, Goldman Sachs Chief Information Officer Marco Argenti had spoken positively about vibe coding—the practice of building software through text prompts without writing manual code, noting how asset management staff could quickly assemble prototypes.

However, the bank is now shifting its approach. Goldman Sachs representative Nick Carraterra explained that the policy revision is based on multiple factors, including operational productivity and risk management. He emphasized that the experimental phase has officially concluded and access will remain exclusively with the personnel best suited for long-term utilization.

Risk management in the financial sector

The decision generated some pushback among employees who had grown accustomed to solving daily operational tasks with AI assistance. Nevertheless, large financial institutions must prioritize strict cybersecurity standards, regulatory compliance, and data governance over casual developer experimentation.

This move highlights the broader tension within corporate environments regarding generative AI. While tools promise unprecedented efficiency, major global corporations remain cautious about the security risks and code quality issues associated with widespread non-technical deployment.