Goldman Sachs Restricts Claude Code, GitHub Copilot to Engineers Citing High Costs and Risk Management

Goldman Sachs has restricted access to advanced AI coding tools like Anthropic's Claude Code and Microsoft's GitHub Copilot, limiting their use to engineers only, according to a Bloomberg report from October. This move reverses a previous stance where non-technical staff, including traders and bankers, had used these tools for months, citing concerns over productivity and risk management. For broader context, explore our AI Tools Pricing.

Goldman Sachs Reverses Broad AI Tool Access

The new internal guidelines from Goldman Sachs, implemented in October, specifically target advanced AI coding tools. Previously, these tools were accessible to a wider range of employees. In July, Goldman CIO Marco Argenti had publicly endorsed a concept of "vibe coding" by non-technical personnel, suggesting a more inclusive approach to AI tool adoption. However, the recent restrictions mean that tools like Claude Code and GitHub Copilot are now reserved for employees "who are best suited to use them for the long term," according to a Goldman spokesperson.

This policy adjustment has reportedly caused frustration among some staff who had integrated these AI tools into their daily workflows for several months. The bank's decision reflects a re-evaluation of its AI strategy, moving from broad experimentation to a more controlled deployment focused on specific technical roles.

Rising Costs and Governance Challenges of AI Coding

The move by Goldman Sachs highlights growing industry concerns regarding the cost and governance of advanced AI coding tools. While AI offers significant potential for productivity gains, the computational resources required for certain applications can be substantial. MIT Technology Review reported that a single agentic coding task can consume 10 to 50 times more tokens than a standard chat prompt, indicating a higher operational cost for complex AI-driven development.

These high token costs are becoming a critical factor for enterprises. For instance, Uber reportedly exhausted its entire 2026 AI budget by April after expanding its usage of Claude Code. This rapid expenditure underscores the financial implications of widespread AI coding tool adoption without stringent cost management. Gartner further predicts that AI coding token costs could surpass average developer salaries by 2028, signaling a significant shift in IT budgeting and resource allocation.

Industry Implications and Future Outlook

Goldman Sachs's decision to limit AI coding tools to engineers suggests a broader trend in the financial sector and beyond. Companies are increasingly balancing the benefits of AI-driven productivity with the practical challenges of cost, governance, and risk management. The initial enthusiasm for "AI coding for everyone" appears to be evolving into a more strategic and controlled implementation.

This shift emphasizes the need for organizations to carefully evaluate the return on investment for AI tools, particularly those with high operational costs. It also underscores the importance of establishing clear guidelines and governance frameworks to ensure responsible and efficient AI deployment. As AI technology continues to advance, businesses will likely refine their strategies to maximize its benefits while mitigating associated risks and expenditures.

Sources

• Toward a Human-Centered Framework for Evaluating AI...
• Finance-LLMs/README.md at main · kennethleungty/Finance-LLMs · GitHub
• Stop counting tokens, start counting what ships | MIT Technology Review
• Goldman Restricts Traders and Bankers From Using Some AI Tools - Bloomberg