Goldman Sachs has restricted most traders and bankers from using advanced artificial intelligence coding tools, including Anthropic’s Claude Code and Microsoft’s GitHub, following an internal review of productivity and risk management concerns.

The Wall Street bank had allowed employees across its business to use the tools for several months but has now limited access to engineers, according to internal guidance updated in October and seen by Bloomberg. The change has frustrated some traders who had incorporated the technology into their daily work.

According to Bloomberg, employees had used the tools to monitor client trading flows, model potential market movements and changes in positions, and produce written and visual content. The restriction does not apply to so-called strats, employees whose responsibilities combine trading, quantitative research and software engineering.

Nick Carcaterra, a Goldman spokesperson, said: “We make certain AI tools available to our people to determine the best use cases across different parts of the firm.”

“After a period of experimentation where we assess a variety of factors including productivity and risk management, we put them in the hands of those who are best suited to leverage them for the long term,” he added.

The decision marks a change in direction from the bank's earlier encouragement of AI experimentation among non-technical employees. In July, chief information officer Marco Argenti praised the use of so-called vibe coding by traders and bankers, who had used AI tools to develop solutions before passing them to engineers for further development.

Argenti said the bank's engineering workforce, which numbers approximately 12,000, could refine and improve applications created by employees without specialist technical expertise.

However, Goldman has faced concerns about the longer-term implications of widespread AI adoption. In August, partner Chris Churchman warned that increasing reliance on the technology could weaken the analytical capabilities of future generations of bankers, according to Quartz.

The restriction comes as financial institutions reassess how employees access and use increasingly sophisticated AI models. Banks are investing millions of dollars in the technology while seeking to manage associated costs, cybersecurity risks and questions over productivity.

JPMorgan Chase said in July that it wanted employees to select AI models according to the complexity of individual tasks, avoiding expensive models for relatively straightforward work such as summarising analyst reports.

Goldman has developed its own AI assistant, combining models from OpenAI, Anthropic and Alphabet. The bank has previously restricted access to Anthropic's Claude among employees in Hong Kong, introducing location-specific controls earlier this year.

In May, John Waldron, Goldman’s president and chief operating officer, described the bank as a “human assembly line” with considerable potential for automation.

---