Goldman Sachs restricts access to advanced AI coding tools

Goldman Sachs has restricted access to advanced AI-powered coding tools, such as Anthropic’s Claude Code and Microsoft’s GitHub, limiting their use to engineers and quantitative specialists. This shift marks a change in strategy for the investment bank, which previously encouraged broader experimentation with AI prototypes among non-technical staff.

A spokesperson for the firm stated that the company is evaluating AI tools based on productivity and risk management, noting that once experimental phases conclude, the bank allocates tools to those best suited to leverage them long-term. This move has reportedly met with resistance from traders who relied on these tools for data monitoring and financial modeling.

The banking sector is currently navigating the tension between rapid AI adoption and operational safety. Leaders at competitors, including JPMorgan Chase and Bank of America, have highlighted the difficulty of managing real-time risks, particularly the danger of AI-generated content lacking sufficient human verification.

Despite the new restrictions, Goldman Sachs continues to pilot internal AI initiatives. The firm currently maintains proprietary chatbots for general staff and remains in the testing phase for autonomous coding agents and research-assistance tools, reflecting an ongoing effort to integrate AI within a controlled corporate framework.